endogenous networks in investment syndication
TRANSCRIPT
Electronic copy available at: http://ssrn.com/abstract=1885299
Forthcoming in Journal of Corporate Finance
Endogenous Networks in Investment Syndication*
Lanfang Wang1 and Susheng Wang2
January 2012
Abstract: As an effective investment strategy, investors often invest jointly in a company by
forming a syndicate. The unique feature of this paper is that it endogenizes the formation of
an investment syndicate. We provide a theory on the endogenous formation of networks in
investment syndication and analyze how several key factors such as risk aversion, productiv-
ity, risk and cost affect incentive and syndicated investment. We also apply the theory to
venture capital investment and identify empirical evidence in support of it.
Keywords: investment syndication, endogenous network, investment risk, risk aversion,
project productivity
JEL Classification: G30
* We gratefully acknowledge helpful comments and suggestions from an anonymous referee, which led to a substantial improvement of the paper.
1 Assistant Professor at Institute of Accounting and Finance, Shanghai University of Finance and Economics.
Email: [email protected].
2 Professor at Hong Kong University of Science and Technology. Email: [email protected].
Electronic copy available at: http://ssrn.com/abstract=1885299
Page 2 of 48
1. Introduction
Risk-averse investors demand large compensation for investing in highly risky projects.
One effective strategy for investors to deal with risks is to form a group to invest jointly in a
firm. This way a number of investors can share the risks, especially when investing in a
highly risky firm. This group of investors is referred to as an investment syndicate and the
company being invested in is referred to as a portfolio company. An investment syndicate is
a form of interfirm alliances in which two or more venture capital firms or investment banks
(investors) co-invest in a portfolio company and share a payoff (Wright and Lockett, 2003).
Investment syndicates are very popular in reality. For example, Wright and Lockett (2003)
reported that 63.6% of U.S. and 29.5% of European venture capital (VC) investments were
syndicated in 2000. Brander et al. (2002) found that 60% of Canadian VC investments in
1997 involved more than one VC investor. Lerner (1994) suggested that syndication is com-
monplace even in first-round investments. In practice, an investment syndicate is typically
led by one of the investors, who actively invites other investors to join in and coordinates
investment deals between the company and the investors. This lead investor typically makes
the largest investment among the group and is involved throughout the whole investment
phase.
In the beginning, the entrepreneur (EN) of the portfolio company initiates an invest-
ment syndicate by inviting and appointing a lead investor. When deciding on whom to ap-
point as the lead investor, the EN must take into account two major factors: (1) the lead
investor’s ability in organizing a syndicate to satisfy the needs of the company; and (2) the
bargaining power of this syndicate for a share of company profits. One crucial determinant of
the lead investor’s ability to organize a syndicate is her business connections, which deter-
mines her ability in inviting suitable investors to join her syndicate. Such personal business
connections are known as the lead investor’s network capital. A well organized syndicate can
provide the necessary funding for the company. At the same time it is also in a good position
to bargain for a large share of company profits. For example, Hochberg et al. (2010) argued
for the value of networks in restricting the entry of outside venture capitalists (VCs), thus
improving the incumbents’ bargaining power with the EN. Hence, when choosing a suitable
lead investor, the EN needs to strike a balance between a syndicate’s ability in providing
funding to satisfy the company’s needs and its potential to bargain for company profits. For
this reason, an equilibrium theory should be developed on the level of the lead investor’s
network capital or more generally on the composition of an investment syndicate in equilib-
rium.
Page 3 of 48
Although syndication is very important in venture capital, little is known about the mo-
tives of syndication. Lerner (1994) pointed out that “Syndication has been little scrutinized in
the corporate finance literature. The reason may lie in the difficulty of analyzing syndication
patterns empirically and the complexity of motives behind syndication.” More recently,
Brander et al. (2002) claimed that “Aside from Lerner (1994), we are unaware of any pub-
lished academic papers in which the rationale for VC syndication is the central focus, al-
though some of the existing literature on VC finance does bear on the syndication question”.
In this study, we provide a theory on the endogenous formation of an investment syndicate.
The theoretical model has three stages. In the first stage, the EN of the portfolio company
selects a lead investor. In the second stage, the lead investor coordinates the formation of a
syndicate. In the third stage, the EN and the syndicate or the lead investor bargain for output
shares. Most of the existing studies have focused on the second stage. Our analysis shows the
impact of network endogeneity. In particular, we show opposite effects of productivity on
effort if a network is treated as exogenous instead of endogenous.
This study is related to two streams of literature. The first stream is on investment syn-
dication. Existing studies on investment syndication have focused on two questions: what is
the rationale behind investment syndication and how are investment syndicates formed?
There are many theoretical studies on the rationale behind syndication. In theoretical studies,
Wilson (1968), Sah and Stiglitz (1986) and Pichler and Wilhelm (2001) suggested that syndi-
cation is an efficient way to share risks among partners and gather additional information
about a startup’s value and can also be a way of colluding tacitly. Admati and Pleiderer (1994)
provided an implicit rationale for syndication based on information asymmetry between VCs
and the EN. They suggested that the only optimal financial contract that can mitigate overin-
vestment and other agency problems is a fixed-fraction contract. This is when the VCs re-
ceive a fixed fraction of the project’s payoff and finance the same fraction of future invest-
ments. Their theoretical model implies that the VCs’ syndication decision in later rounds
would ensure the optimality of a financial contract. Casamatta and Haritchabalet (2007)
suggested that VC syndication improves the screening process of VCs and prevents competi-
tion among investors after investment opportunities are disclosed. They also highlighted the
importance of experience to the formation and efficiency of investment syndicates, which
suggests that experienced VCs should syndicate with experienced VCs. Huang and Xu (2003)
provided a contractual foundation that solves a class of commitment problems in R&D fi-
nancing and argued that syndicated VC financing can be deployed as a commitment device to
terminate bad projects promptly. Fluck et al. (2006) provided an explanation to the popular
co-existing phenomena of staged financing and syndication in VC investments. They found
that ex ante commitment to syndicate investments in later stages protects the EN from po-
Page 4 of 48
tential hold-up problems arising from staged financing. Dorobantu (2006) suggested that a
VC can use syndication in the second or a later financing round as a signaling device to
credibly communicate private information about her project-selection ability to potential
investors in a follow-on round.
The most related empirical works on the rationale of investment syndication are those of
Brander et al. (2002) and Ozdemir (2006). Brander et al. (2002) examined two conjectures:
the venture selection and value-added hypotheses. Using Canadian VC investment data, they
found that syndicated VC investments have higher returns, which supports the value-added
interpretation. A similar result was found by Cumming and Walz (2010). Ozdemir (2006)
discussed “how social structure and a VC firm’s position within it affect the propensity of
syndication.” Ozdemir confirmed that “uncertainty and risk associated with investing in a
particular target company increase the likelihood of syndication.” De Clercq and Dimov
(2004) investigated the realized investment strategies of 200 U.S.-based VCs over a twelve-
year period. Their findings suggested that syndication is a mechanism for VCs to share com-
plementary knowledge and financial risks among syndicate partners.
Unlike the rationale behind investment syndication, there are few theoretical studies on
the formation of investment syndicates. To our knowledge, the only related studies to our
work are those of Cestone et al. (2007) and Tykvová (2007). Cestone et al. (2007) focused on
contract design to induce VCs to truthfully disclose investment opportunities. Their findings
suggest that more experienced lead VCs should select more experienced partners. This con-
clusion is consistent with the implications of the theoretical work of Casamatta and Haritch-
abalet (2007) and is supported by the empirical works of Lerner (1994) and Du (2009). In
contrast, Tykvová (2007) showed in a complete information setting that VCs may choose to
syndicate with less experienced partners if the latter are willing to accept comparably worse
payoffs after taking into account the benefits of reputation building and the transfer of know-
how among partners.
The second stream of literature related to this study consists of recent studies on the ap-
plication of network theory in finance, especially VC investments. Network structure is
known to be important in determining the outcome of many important social and economic
relations.3 In finance, network theory is mainly applied to financial stability and contagion.
3 Examples of the effects of social networks on economic activity are abundant, including their effects on
transmitting information about job availability, new products, technologies and political opinions. They also
serve as channels for informal insurance and risk sharing. The network structure influences patterns of decisions
regarding education, career, hobbies, criminal activity and even participation in micro-finance. Durlauf (2004)
Page 5 of 48
Freixas et al. (2000), Leitner (2005), Dasgupta (2004) and Cummins et al. (2002) showed
that the existence of more network connections among banks may reduce the risk of conta-
gion. Eisenberg and Noe (2001), Afonso and Shin (2007) and Gai and Kapadia (2007) used
the techniques developed in mathematics and physics to study contagion. Lagunoff and
Schreft (2001) and Cifuentes et al. (2005) studied the impact of indirect linkages on conta-
gion. Babus (2009), Allen and Gale (2000) and Castiglionesi and Navarro (2008) studied the
issue of network formation among banks. Also, Babus’ (2009) theoretical study on the risk of
contagion among banks suggested endogenous network formation among banks. She
showed that contagion does not occur in equilibrium.
In empirical studies, Ozdemir (2006) showed that “those VC firms deeply embedded in
the VC industry syndication and social networks are more likely to initiate syndicates, al-
though the more experienced VC firms prefer going solo.” Hochberg et al. (2007) investi-
gated the effect of networks on VC investments. They looked at VCs who are connected
through a network of syndicated investments. They found that better-networked VCs experi-
ence significantly better fund performance, as measured by the proportion of investments
that are successfully exited through IPOs and sales to other companies. Hochberg et al.
(2010) further showed that VC syndication presents a potential barrier to entry for new VCs,
and incumbents appear to have benefited from this reduced entry because they show the
willingness to accept lower compensation for their deals. Hochberg et al.’s (2007, 2010)
works agrees with the earlier work of Seppä and Jääskeläinen (2003), who explored 54,700
VC investments in 10,057 portfolio companies by 100 of the largest private U.S. VCs during
1986–2000. Sorenson and Stuart (2001) explored how interfirm networks in the U.S. VC
market affect spatial patterns of exchange. Empirical evidence suggests that social networks
in the VC market—built up through the extensive use of syndicated investments—transfer
information across boundaries and therefore expand the spatial radius of exchange.
Sorenson and Stuart (2008) proposed a theory to explain the formation of distant ties in VC
investment networks and suggested that VCs form relations with distant partners when they
participate in two types of settings: unusually faddish ones and those with limited risks to
participants.
This paper examines a different aspect of the formation of investment syndicates. Previ-
ous theoretical and empirical studies have emphasized relational ties among investors but
reviewed theoretical models on neighborhood effects. Allen and Babus (2009) reviewed the application of net-
work theory in finance. Jackson (2005, 2010) gave a survey of theoretical works on network formation and
provided an overview of social networks in economic applications.
Page 6 of 48
ignored the role of the EN. In other words, they have focused on how the lead investor
chooses syndication partners, with the EN playing no role in the process. We, however, take
into account the role of the EN in the formation of the investment syndicate. We propose a
theoretical model to explain how the syndicate is determined through a process involving
choices and bargaining between the EN and the syndicate and among the syndicate members.
In our theoretical model, the syndication process involves three stages: the EN first chooses
the lead investor; then the lead investor chooses syndication partners; and finally the EN and
the syndicate bargain to determine their output shares. The first and final stages have so far
been ignored in the existing literature on the formation of investment syndicates. Our three-
stage syndication process endogenously determines the lead investor’s network capital in
equilibrium.
This paper contributes to the literature on the networks of investment syndicates. Our
model suggests that when choosing the lead investor, the EN will take into account the lead
investor’s network capability as a key factor. This is quite consistent with existing empirical
evidence. For example, Gulati (1999) found that accumulated network resources (or what we
refer to as network capital in this paper) arising from the history of participation in the net-
work of alliances are influential in a firm’s decision to join new alliances. More direct evi-
dence came from Smith (2001), who found that the factors the EN considers when evaluating
VCs include network capital, such as “co-investing with other investors”, “interfacing with
the investor group” and “obtaining alternate sources of financing.” A lead investor’s network
capital affects her ability to organize a sizable syndicate, which is proven to be valuable to
project efficiency (Brander et al., 2002). We conclude that the tradeoff between syndicated
funding and profit sharing determines the optimal choice of the EN, by which the lead inves-
tor’s network capital is endogenously determined in the process. Endogenously determined
networks can be found in the existing studies on financial networks. For example, Babus
(2009) found that a financial network of banks emerges endogenously. In equilibrium, with
endogenous networks, the degree of systemic risk is significantly reduced. Further, in certain
equilibria, contagion does not occur.
Our theoretical model shows that the determinants of the syndicate size or the lead in-
vestor’s network capital include the investors’ risk aversion, project productivity and inves-
tors’ output share. We further provide empirical evidence to support our theoretical findings
using 40,395 identified domestic VC investment rounds made in 15,264 portfolio companies
by U.S. VCs during 1985–2005. After controlling for the type of lead VC firm (lead VC type)
and the fixed effects of funding year, industry and round sequence, we found that investment
risk and portfolio company quality had a significant positive effect on the syndicate size. This
empirical finding is consistent with our theoretical results.
Page 7 of 48
This paper is organized as follows. In Section 2, we set up the model. In Section 3, we
solve for the equilibrium solution and provide a theoretical analysis of our solution. In Sec-
tion 4, we apply our theory to VC investments and present an empirical analysis using data
from the U.S. VC industry. We conclude in Section 5 with a summary of our main results and
remarks. The derivations are given in the Appendix.
2. The Model
In this section, we develop a theory on the formation of an investment syndicate supply-
ing funds to a target company.
Consider an EN who seeks investment from investors for a project. Funding is provided
through a syndicate. The syndicate is organized by a lead investor, who is invited and ap-
pointed by the EN. The syndicate is defined by a vector 1( , , , ),nn a a where n is the num-
ber of investors in the syndicate and is called the lead investor’s network capital, and ia is a
measure of the risk aversion of investor i in the syndicate. This n is endogenous. For the
investors, a larger n implies more investors to share risk and increases the bargaining power
of the syndicate for a share of output; for the EN, a larger n means more funding but re-
duces the EN’s bargaining power. An optimal n will be chosen by the EN when she selects
the lead investor.
The EN is risk neutral, and the investors are risk averse. Each investor has utility func-
tion:
11( ) .1
ii
i
u x x a
a-=
- (1)
Output is determined jointly by investment from investors and effort from the EN. The EN
has an incentive problem since her effort a is unverifiable. Specifically, output ex ante is a
lottery x defined by
( )( ), ( ); 0,1 ( ) ,x x I p a p aº -
where I is the total investment from the investors, ( )x I is the output ex post when the
project is successful, ( )p a is the probability of success, and both ( )x I and ( )p a are as-
sumed to be increasing functions. We assume ( ) ( )p a x I I* * *> in equilibrium ( , ),a I* *
implying a profitable project.
The EN bargains with the lead investor over how to divide the output after it is pro-
duced. Let ( )nl be the output share for the syndicate. Hence, ( )nl represents the bargain-
Page 8 of 48
ing power of the lead investor. We assume that ( )nl is increasing in ,n meaning that the
larger the syndicate, the more bargaining power the lead investor has.
We take the same incomplete contract approach as Hart and Moore (1990) did. In our
model setting, there is no contract ex ante at 0t = and the two parties, the EN and the syn-
dicate (SN), bargain for output shares ex post at 1.t = However, the equilibrium solution is
implementable by a complete contract ex ante. If n* is the equilibrium solution, the com-
plete contract is an upfront equity-sharing agreement that gives an equity share of ( )nl * to
the syndicate and an equity share of 1 ( )nl *- to the EN. Our model setting ensures our
solution to be bargaining-proof or renegotiation-proof.
We solve the problem backwards in three steps as indicated by the following figure.
0 1
Investors jointly provideEN applies effort
Ia
Bargain for output, for SN, 1 for ENl l-EN chooses n
Step 1: The EN’s Effort
After output is produced, the two parties bargain for a share. We assume that when bar-
gaining for a share of output, the lead investor representing the syndicate behaves as a risk-
neutral agent. This is symmetric to the setting on the EN’s side, where the risk-neutral EN
also represents a group of agents in reality, including managers, entrepreneurs and the
original owners (insiders) of the firm. Only when each investor considers her own invest-
ment does she behave as a risk-averse agent. We use the Nash bargaining solution to define
their output shares. If no agreement is made, both parties get nothing; if there is an agree-
ment, they share output .x The Nash bargaining solution implies the following ex post pay-
offs for the EN and the syndicate, respectively,
[ ]1 ( ) , ( ) ,EN SNn x n xl lP = - P =
where the subscript SN stands for the syndicate, random output x takes either ( )x I or 0,
and ( )nl represents the lead investor’s bargaining power, which is assumed to be increasing
in .n With private cost ( )c a and risk neutrality, the EN’s ex ante payoff is4
[ ]1 ( ) ( ) ( ) ( ),ENU n p a x I c al= - -
4 We can also assume that the EN is risk averse, which does not add much complexity to our analysis nor
change our conclusions.
Page 9 of 48
where ( )c a is the EN’s private cost of effort. Then, the EN chooses a according to the fol-
lowing first-order condition (FOC):
[ ]1 ( ) ( ) ( ) ( ).n p a x I c al ¢ ¢- = (2)
By assuming concave ( )p a and convex ( ),c a the second-order condition (SOC) for the
choice of a is satisfied.
Step 2: The Investors’ Investment
Consistent with reality, assume that the expected return is the same among all the inves-
tors in a syndicate. This means that each investor’s income from investment is proportional
to her investment. That is, investor i with investment iI receives a share /iI I from the
syndicate’s total income ( ) .n xl With the coordination of the lead investor, the investors
form a cooperative group. Suppose that the total investment I is to be determined by social
welfare maximization in the syndicate. Then, assuming (0) 0,iu = the syndicate’s investment
decision is determined by the following problem:
( ){ } 1
max ( ) ( ) .i
ni
SN i j ijI i jj
IU p a u n x I II
l=
ì üé ùï ïï ïê úº -í ýê úï ïê úï ïë ûî þå åå
This problem implies the total investment .I That is, it is not the lead investor who decides
;I instead, it is the syndicate that determines this I together. Based on the existing litera-
ture, we assume homogeneity among the investors in a syndicate; specifically, we assume
that all the investors in a syndicate have the same risk aversion .a Hence, we must have
/iI I n= for all i in equilibrium. Then, the FOC for the above problem is (see the Appendix
for the derivations):
1( ) ( ) ( )1 ( ) .
( )n x I x Ip a nn x I
al - ¢é ùê ú=ê úë û
(3)
This equation determines the optimal funding ˆ( , ).I a n That is, the investment is affected by
the EN’s incentive and the lead investor’s network capital.
Our assumption of homogenous investors is well justified by the literature. Casamatta
and Haritchabalet (2007) studied the rationale behind syndication. They indicated that
syndication can improve the screening process (gather information) and prevent competition
among investors. They highlighted the importance of experience to the formation and effi-
ciency of a syndicate and suggested that experienced VCs syndicate exclusively with experi-
enced partners because they possess good assessment skills. Cestone, Lerner, and White
(2007) studied the formation of VC syndicates. They also found that more experienced VCs
Page 10 of 48
select more experienced partners. Du (2009) examined VCs’ preferences for syndication
partners in an empirical study. She found that VCs are less likely to syndicate with partners
who are different from themselves.
Step 3: The EN’s Choice of Network Capital
The EN selects a lead investor with a certain network capital .n Given equations (2) and
(3), the EN’s problem is
[ ]
[ ], ,
1
1
2
max 1 ( ) ( ) ( ) ( )
s t : 1 ( ) ( ) ( ) ( ),
( ) ( ) ( ): ( ) 1.( )
n a In p a x I c a
IC n p a x I c a
n x I x IIC p a nn x I
a
l
l
l -
- -
¢ ¢. . - =
é ù ¢ê ú =ê úë û
(4)
Notice that we do not include individual rationality (IR) conditions in the problem, which
can be expressed as
[ ] ( )1
1 ( ) ( ) ( ) ( ) 0, ( ) ( ) 0.n
ii j ij
i jj
In p a x I c a p a u n x I II
l l=
ì üé ùï ïï ïê ú- - > - >í ýê úï ïê úï ïë ûî þå åå
We have no need for the IR conditions in problem (4) since we have implicitly allowed the
bargaining process to include a monetary transfer between the two parties to ensure both IR
conditions are satisfied for the EN and the syndicate. It turns out that such a transfer is
unnecessary in our model since for the set of functions we choose later in (5), the solution of
(4) automatically satisfies the two IR conditions.
To gather a group of investors to satisfy a company’s needs for funding, the lead investor
must have many network connections in the investor community. On the one hand, a well-
connected investor is able to gather a proper and sufficiently large group of investors to
provide sufficient funding and to spread risks among the investors. On the other hand, such
a well-connected lead investor has more bargaining power against the portfolio company.
Through Nash bargaining, a well-connected investor obtains a large income share from a
deal, which reduces the EN’s incentive to work. Hence, the EN needs to balance between
these two factors when she appoints a lead investor. The optimal network capacity of this
lead investor is the key in our model.
Most existing theoretical studies on investment syndication have focused on how a lead
investor chooses partners but ignored the role of the EN. Our model lets the EN choose a
lead investor with a certain network capital. This is consistent with the study by Smith
(2001), whose survey data suggests that companies in need of investment actively choose
Page 11 of 48
investors. He found a number of network factors to be important in this choice, including
“co-investing with other venture capitalists,” “interfacing with the investor group” and “ob-
taining alternate sources of financing.”
Theoretical and empirical studies have suggested that both the EN and the lead investor
are important to the project. In existing theoretical studies, either the EN or the investors
have been assumed to be the principal in their relationship. For example, in a study by Ad-
mati and Pfleiderer (1994), the EN determines the initial financing agreement in order to
maximize the net surplus of the project, while in another study by Amit et al. (1990) the VCs
determine the manager’s profit share in order to maximize their own profits. We assume that
the EN chooses a lead investor to initiate the syndication process, which is quite consistent
with the empirical evidence in the survey by Smith (2001). Smith (2001) indicated that
around 71% of ENs in his sample had a choice on which VCs should be allowed to invest in
their companies.
3. Theoretical Analysis
3.1. The Cooperative Investment Solution
To analyze the solution, we must use parametric functions. Parametric functions allow a
few parameters to represent the key aspects of the issue at hand. We choose the following
parametric functions:5
11( ) , ( ) , ( ) , ( ) , ( ) ,1iu x x p a a c a a x I I n na g b d l r
a-= = = = =
- (5)
where , , , (0, 1)a g d r Î and 1.b ³ These parameters represent the key factors of the prob-
lem and allow us to analyze various aspects of the solution. Here, a is a measure of risk
aversion, d represents the productivity of investment (effective use of investment), and r is
an investor’s output share. For our equilibrium solution ( , , ),n a I* * * since 1a* < with rea-
sonable parameter values, an increase in b tends to reduce the cost of effort ( ).c a* Hence,
b represents easiness of effort. Similarly, an increase in g tends to reduce the chance of
5 The results do not change if we use the following more general functions:
1( ) , ( ) , ( ) , ( ) , ( ) ,iu x Ax p a Ba c a Ca x I DI n E na g b d l r-= = = = =
where , , ,A B C D and E are arbitrary constants.
Page 12 of 48
success ( ).p a* Hence, g represents project riskiness. Since the ratio /g b often appears in
our formulas, we denote / .q g bº
Given the above parametric functions, we can solve for a closed-form solution of prob-
lem (4). The derivations are in the Appendix. The solution of the network capital is:
.(1 )
n dr ad
* =+
(6)
Then,
2 ( )(1 )( )(1 )
12 ( )(1 )( )(1 )
1 ,1 1
1 .1 1
I
a
b ggb g ad bdb g ad bd a
dad db g ad bdb g ad bd a
ad d g dr
ad b ad
ad d g dr
ad b ad
-- + -- + -* -
+ -- + -- + -* -
æ öæ ö+ - ÷ç÷ç ÷= ÷ çç ÷÷ ÷ç÷ç è ø+ +è ø
æ öæ ö+ - ÷ç÷ç ÷= ÷ çç ÷÷ ÷ç÷ç è ø+ +è ø
(7)
The syndicate’s output share is
.1
dl
ad* =
+ (8)
In contrast, if the network capital is exogenous (i.e., n is a given constant), with a corre-
sponding change to problem (4), the solutions of effort and investment are
( ) ( )
( ) ( )
( )(1 ) 1 ( )(1 )
1( )(1 ) 1 ( )(1 )
ˆ 1 ,
ˆ 1 .
I n n
a n n
gb gb g ad bd a b g ad bd
ad ddb g ad bd a b g ad bd
gr dr
b
gr dr
b
-- + - - - + -
+ -- + - - - + -
é ùê ú= -ê úë û
é ùê ú= -ê úë û
(9)
3.2. The Nash Investment Solution
In the above solution, the investors in the syndicate decide their investments coopera-
tively. Hence, we call the above solution the cooperative investment solution. Alternatively,
the investors in the syndicate may play a Nash game to determine their investments.6 The
solution in this case is called the Nash investment solution and investor i ’s investment is
determined by
6 We would like to acknowledge and thank an anonymous referee for proposing this alternative solution and
for suggesting that we compare syndicated investment with solo investment.
Page 13 of 48
( )max ( ) ( ) .i
ii i j ijI
jj
IU p a u n x I II
lé ùê úº -ê úê úë û
åå
The FOC is
1( ) ( ) 1 ( )1 ( ) .
( )n x I n x Ip an I x I
al - é ùé ù ¢-ê úê ú= +ê úê úë û ë û
(10)
Then, the EN’s problem becomes
[ ]
[ ], ,
1
1
2
max 1 ( ) ( ) ( ) ( )
s t : 1 ( ) ( ) ( ) ( ),
( ) ( ) 1 ( ): ( ) 1.( )
n a In p a x I c a
IC n p a x I c a
n x I n x IIC p an I x I
a
l
l
l -
- -
¢ ¢. . - =
é ùé ù ¢-ê úê ú + =ê úê úë û ë û
(11)
Its solution is
(1 )(1 ) .
(1 )n d r d ad d
r ad* + - + -
=+
Then,
( )
( )
( )(1 ) 1 ( )(1 )
1( )(1 ) 1 ( )(1 )
(1 ) 1 .
(1 ) 1 .
I n n
a n n
gb g
b g ad bd a b g ad bd
d add
b g ad bd a b g ad bd
gr r d
b
gr r d
b
-- + -* * - * - + -
- +- + -* * - * - + -
é ù é ùê ú= - + -ë ûê úë û
é ù é ùê ú= - + -ë ûê úë û
If the network capital is exogenous, with a corresponding change to problem (11), the solu-
tions of effort and investment are
( )
( )
( )(1 ) 1 ( )(1 )
1( )(1 ) 1 ( )(1 )
ˆ (1 ) 1 ,
ˆ (1 ) 1 .
I n n
a n n
gb gb g ad bd a b g ad bd
ad ddb g ad bd a b g ad bd
gr r d
b
gr r d
b
-- + - - - + -
+ -- + - - - + -
é ù é ùê ú= - + -ë ûê úë û
é ù é ùê ú= - + -ë ûê úë û
3.3. Analysis
We analyze the solutions in this section. It turns out that the results derived from the
two alternative solutions are very similar. Hence, we present the results for the cooperative
Page 14 of 48
investment solution only. Due to the space limit, we have only included the proof of Proposi-
tion 1 in the Appendix; other proofs are available upon request.
Syndication versus Solo
Is it possible to have a syndicated investment that is smaller than solo investment (i.e.
underinvestment)? Is it possible that incentive is worse under syndication? By denoting I
and a in (9) respectively as functions ˆ( )I n and ˆ( )a n of ,n we can compare solo investment ˆ(1)I with syndicated investment ˆ( ),I I n* *= and compare effort ˆ(1)a under solo investment
with effort ˆ( )a a n* *= under syndication. The question is: is it possible to have ˆ ˆ( ) (1)I n I* <
and ˆ ˆ( ) (1)a n a* < when 1?n* ³ It turns out that it is.
Proposition 1 (Syndication vs. Solo). Given c0ndition 1,n* ³ i.e., ,1
dr
ad£
+ for the
cooperative investment solution,
(a) Solo investment is larger than syndicated investment if and only if
1 ,1
dr q
ad£ - £
+ (12)
where / .q g bº
(b) Incentive under solo investment is better if and only if
1 .1
dq
ad- £
+ (13)
For example, if 0.1,a = 1.5,b = 0.8,g = 0.2r = and 0.8,d = the cooperative solu-
tion indicates underinvestment with syndication: 3.7,n* = ˆ( ) 4.9I n* = and ˆ(1) 5.1.I =
Similarly, for the same parameter values, the Nash investment solution also indicates under-
investment with syndication: 3.8,n* = ˆ( ) 50.5I n* = and ˆ(1) 64.7.I =
There are two inequality conditions in (12):
1 and 1 .1
dr q q
ad£ - - £
+
If and only if one of these inequality conditions fails will syndicated investment be larger.
The second inequality condition is the same as (13). This means that if the first inequality
condition holds, the EN’s incentive and investors’ investment are strategic complements in
equilibrium with respect to syndication; if the first inequality condition fails but the second
one holds, incentive and investment are strategic substitutes. In the latter case, solo invest-
Page 15 of 48
ment is smaller, but incentive under solo investment is better. In the following, we provide
intuition for the results in Proposition 1. We focus on investment only; intuition for incentive
is similar.
On the one hand, with a solo investor, although sources of funding are limited, her in-
centive is good since she is the sole recipient of the income share for investors and hence
may invest a lot. On the other hand, with syndication, although there are multiple sources of
funding, each investor’s willingness to invest is low since the investors have to share the
benefits and hence the total investment may be low. Both the EN and the investors care
about output shares, but while the EN welcomes more investment, the investors prefer to
invest less. Depending on project risk, risk aversion, productivity and cost of effort, each
party may have a tendency towards one objective or the other. In cases where the EN does
not care so much about the investment amount, for example, if productivity is high (in terms
of effective use of funding), solo investment may be larger than syndicated investment. Our
theory indeed suggests this. Condition (12) is more likely to hold if d is large; if so, solo
investment is larger than syndicated investment.
One important function of the syndicate is to share risk. If risk aversion is low, the syn-
dicate’s role of risk sharing is reduced, while output sharing within the syndicate will still
have a negative effect on investment. Hence, if risk aversion is low, solo investment may be
larger than syndicated investment. Our theory indeed confirms this. Condition (12) is more
likely to hold if a is small; if so, solo investment is larger than syndicated investment. On
the other hand, if risk aversion is high, a solo investor will invest conservatively, while syndi-
cation allows risk sharing so that investors will be more willing to invest. In this case, we
expect syndicated investment to be larger than solo investment. Indeed, when a is large,
condition (12) is more likely to fail; if so, our theory predicts that syndicated investment is
larger.
If each investor only has a small output share ,r each will invest a small amount. Then,
for a given required amount of investment, the syndicate will need to be large. Such a syndi-
cate tends to bargain for a large share of output, causing a negative effect on incentive, which
in turn reduces investment. Hence, in this case, solo investment is likely to be larger than
syndicated investment. Our theory indeed confirms this. Condition (12) is more likely to hold
if r is small; if so, solo investment is larger than syndicated investment.
When the project is highly risky, we expect a preference for syndication on both sides.
Indeed, when the project is risky as represented by a large ,g condition 1r q£ - in (12)
tends to fail, implying a larger syndicated investment.
Page 16 of 48
If effort is costly, the EN will rely more on investment than her own effort, in which case
she is likely to go for syndication. Indeed, if b is small, condition 1r q£ - in (12) tends to
fail; if so, syndicated investment is larger.
Proposition 2 (The Gap between Syndicated and Solo Investments). Given c0ndition
1,n* ³ for the cooperative investment solution,
(a) If 11
dq
ad- >
+or 1 ,
1d
r qad
£ - £+
higher productivity implies a larger gap be-
tween syndicated investment and solo investment.
(b) If 11
dq
ad- >
+ or 1 ,
1d
r qad
£ - £+
higher risk aversion implies a smaller gap
between syndicated investment and solo investment.
(c) If 1 ,r q³ - a larger r implies a larger gap between syndicated investment and solo investment; otherwise the gap is smaller.
Network Capital
Interestingly, g and b are irrelevant to the equilibrium network capital, even though
the equilibrium investment and effort are dependent on these two factors. Here, b repre-
sents easiness of effort and g represents project riskiness, both of which depend crucially on
the EN’s effort. This suggests that those factors directly related to the EN’s effort do not have
an effect on network capital.
In contrast, the network capital is heavily dependent on the risk aversion a and produc-
tivity d of investment, both of which depend crucially on investors and investment. This
suggests that those factors directly related to investment influence network capital strongly.
Proposition 3 (Network Capital).
(a) The syndicate size or network capital is diminishing in risk aversion, and this negative
effect of risk aversion on syndicate size is also diminishing in risk aversion:
2
20, 0.n na a
* *¶ ¶< >
¶ ¶ (14)
(b) The syndicate size or network capital is increasing in productivity, and this positive
effect of productivity on syndicate size is also diminishing in productivity:
Page 17 of 48
2
20, 0.n nd d
* *¶ ¶> <
¶ ¶ (15)
Proposition 3 suggests that more risk aversion implies a smaller syndicate. This result
makes sense since highly risk-averse investors demand high compensation for their invest-
ment. Consequently, the EN prefers a small syndicate to reduce the investors’ bargaining
power for compensation. Proposition 3 also indicates that for a project with high return
potential, the EN will pick a lead investor to organize a large syndicate. This result is fairly
intuitive.
Output Sharing
We now investigate how the two parties share output.
Proposition 4 (Sharing). Regarding the syndicate’s output share ,l* we have
2 2
2 20, 0; 0, 0.l l l la a d d
* * * *¶ ¶ ¶ ¶< > > <
¶ ¶ ¶ ¶ (16)
That is, the more risk averse the investors are, the smaller the output share they will collec-
tively get; and the more productive the investment is, the larger the output share the inves-
tors will collectively get.
A smaller output share for more risk-averse investors is a consequence of less bargaining
power implied by (14). At the same time, a larger output share for the EN induces the EN to
invest more effort. This suggests that providing better incentive for the EN is an effective
strategy to cope with investor risk aversion. Further, if investment is more productive (more
effective use of investment), the investors are given a larger output share, which encourages
them to invest more in the project.
Investment and Incentives
We now investigate the effects of risk aversion, productivity and an investor’s output
share on incentive and investment.
While l is the output share of the syndicate, r is the output share of each investor. For
example, if the syndicate has a 50% share of the output and there are five investors ( 5),n* =
each investor is entitled to 50% / 10%nr *= = of the output. This r reflects each investor’s
bargaining power, which tends to have a negative effect on the EN’s incentive. The effect of
r is clear from (6) and (7): an increase in the investor share r reduces the syndicate size,
Page 18 of 48
effort and investment. The reason is that a larger investor share reduces the EN’s incentive,
which in turn reduces investment; consequently the EN will choose a smaller syndicate size.
The effects of risk aversion and productivity are summarized by the following two
propositions.
Proposition 5 (The Effect of Risk Aversion on Investment and Incentive).
(a) Investment generally increases with risk aversion. Specifically, investment increases
with risk aversion if and only if
( ) ( )[ ] ( )( )
2
111 1 1
2 1 11 111 1 .e
dad
qqad d add q qqr d ad ad d q
æ ö÷ç + - ÷ç ÷çè ø--- -- + - +- --£ + + - (17)
(b) Incentive generally improves with risk aversion. Specifically, incentive improves with
risk aversion if and only if
( ) ( )[ ]1 1 1
21 111 1 .q
d q qqr d ad ad d q- --- --£ + + - (18)
Typically /q g bº is small since b is larger than 1. Hence, we will generally have
1 .d q< - If so, Proposition 5(a) can be expressed as 0Ia
*¶³
¶ if and only if
( ) ( )[ ] ( )( )
2 1 11
1
111
2 1 111.
1 1 ed q
dad
qqad d adqq
rd ad ad d q
--
æ ö÷ç + - ÷ç ÷÷çè ø---- + - +--
ì üï ïï ï£ í ýï ï+ + -ï ïî þ
With reasonable parameter values (where a assumes values between 0 and 1), the right-
hand side of the above inequality is above 70. Hence, condition (17) generally holds. This
means that investment generally increases with risk aversion. By the same reasoning, incen-
tive generally improves with risk aversion.
The explanation is that risk aversion puts pressure on performance, which improves in-
centive and in turn induces more investment. Higher risk aversion also reduces the syndicate
size and hence results in a larger output share for the EN, which also improves incentive.
One interesting observation here is that more risk-averse investors end up investing more in
equilibrium.
Proposition 6 (The Effect of Productivity on Investment and Incentive).
(a) Investment increases with productivity if and only if r is small. Specifically, invest-
ment increases with productivity if and only if
Page 19 of 48
( )( )
11 22 1 1(1 ) (1 ) 1 1 11 .
1 1e
qad q da q qa q
a q a qa d ad d adad d dr q
ad ad
æ öæ ö+ ÷ ÷ç ç- - - -÷ ÷ç ç- ÷ ÷÷ ÷ç çè øè ø- - + + - +æ öæ ö+ - ÷ç÷ç ÷£ ÷ çç ÷÷ ÷ç çè ø è ø+ +
(19)
(b) Incentive is generally decreasing in productivity. Specifically, incentive is decreasing in
productivity if and only if
112 (1 ) (1 )1 .
1 1e
qda da q
a a qd ad dr q
ad ad
+ ---
æ ö æ ö+ -÷ç ÷ç÷³ ÷ç ç÷ ÷÷ç çè ø è ø+ + (20)
For reasonable parameter values, the right-hand side of (19) is between 0 and 10, and
the right-hand side of (20) is less than 0.17. Hence, investment is increasing in productivity
if and only if r is small, and incentive is generally decreasing in productivity. There are two
reasons for these results. On the one hand, since a larger investor output share r has a
negative effect on incentive, r cannot be too large for the EN to have enough incentive. If the
investor share is small, with sufficient incentive, the EN welcomes higher productivity and
applies more effort, which in turn induces investors to invest more. On the other hand, a
large investor share sufficiently dampens the EN’s incentive; in this case, with insufficient
incentive, higher productivity puts less pressure on the EN to perform, which implies less
incentive and investment.
Cost of Effort and Project Risk
We now investigate the effects of easiness of effort and project riskiness on investment
and incentive.
Proposition 7 (The Effects of Easiness of Effort and Project Riskiness).
(a) Investment is generally positively associated with easiness of effort and negatively
associated with project riskiness. Specifically, an increase in easiness of effort b or a
decrease in project riskiness g raises investment if and only if
11(1 )(1 )21 .
1 1e
ad dq ad daad
adad d dr q
ad ad
+ -- + -æ öæ ö+ - ÷ç÷ç ÷³ ÷ çç ÷÷ ÷ç çè ø è ø+ +
(21)
(b) Incentive is generally negatively associated with project riskiness. Specifically, a de-
crease in project riskiness g improves incentive if and only if
11(1 )(1 )12
11 .1 1
ead d
q ad dad daadadq adad d d
r qad ad
+ -- + -+ -
+æ öæ ö+ - ÷ç÷ç ÷³ ÷ çç ÷÷ ÷ç çè ø è ø+ +
(22)
Page 20 of 48
(c) Incentive is generally positively associated with easiness of effort. Specifically, an in-
crease in easiness of effort b improves incentive if and only if (21) holds.
For reasonable parameter values, the right-hand sides of (21) and (22) are less than
0.02 and 0.06, respectively. Hence, these two conditions generally hold. Since 1a* < with
reasonable parameter values, if a* is fixed, an increase in b will reduce the cost of effort
( ).c a* Hence, we expect an increase in b to raise effort, which in turn encourages more
investment. There is a second effect: an increase in effort will raise the cost of effort ( ).c a* In
fact, ( )c a* does increase with b when a* is allowed to change with .b However, this second
effect is a consequence of the first effect. This means that the second effect cannot change the
fact that a* is increased; it can only reduce the extent of this increase. The net effect (the
total of the two effects) is indicated in Proposition 7, i.e., an increase in b raises effort and
investment.
Similarly, since 1,a* < if a* is fixed, an increase in g will reduce the chance ( )p a* of
project success. In fact, ( )p a* is still decreased when a* is allowed to change with .g Hence,
as indicated in Proposition 7, a decrease in g improves the chance of success or reduces
project riskiness, which in turn boosts investment and improves incentive.
So far, we have considered the usual circumstances for the results in Propositions 5–7.
Under some unusual circumstances, when the inequality conditions fail, the conclusions can
be reversed. For example, for Proposition 7(b), if r is sufficiently small, condition (22) fails,
so that incentive is positively associated with project riskiness .g To explain this result, we
need to understand the three underlying forces at work. First, an increase in project riskiness
g has a negative effect on incentive. Second, a small r (a small investor output share) has a
positive effect on incentive. Third, as indicated in Proposition 7(a), an increase in project
riskiness implies less investment, which may put pressure on the EN to perform, implying a
positive effect on incentive. Here, we cite the fact that incentive and investment may be
strategic substitutes in equilibrium. In aggregate, when r is small enough, the positive effect
from a small r and/or from the pressure of reduced investment may dominate, which leads
to an improvement in incentive. Hence, in Proposition 7(b), a positive association between
incentive and project riskiness is possible. Similar situations apply to the other results too.
Page 21 of 48
Endogenous vs Exogenous Networks
One natural question to ask is whether the type of networks, either endogenous or ex-
ogenous, is important in our results. In Figure 1, we graph investment by arbitrarily setting
0.2,r = 0.5,g = 2b = and allowing risk aversion to change; in Figure 2, we regraph
investment with the same parameter values except a fixed 0n at an equilibrium value. Com-
paring Figures 1 and 2, we find that productivity has opposite effects on investment in the
two figures. In other words, with an endogenous network in Figure 1, productivity has a
positive effect on investment; but with an exogenous network in Figure 2, productivity has
a negative effect on investment. Hence, the endogeneity of network is important in our con-
clusions.
0.01
0.02
0.03
0.04
0.24 0.26 0.28 0.3 0.32a
I *
0.3d =
0.5d =
0.06
0.07
0.08
0.09
0.1
0.11
0.24 0.26 0.28 0.3 0.32a
I
0.3d =
0.5d =
Figure 1. Investment with an Endogenous Network Figure 2. Investment with an Exogenous Network
4. Empirical Evidence from VC Investments
Our theory is applicable to many kinds of investments. With the guidance of our theory,
in this section, we conduct an empirical analysis of the determinants of the syndicate size
using U.S. VC investment data.
Syndication is widely observed in VC investments. In fact, VCs typically syndicate their
investments with other VCs, rather than investing alone (Lerner, 1994). The role of syndica-
tion has been investigated theoretically and empirically in the literature. Casamatta and
Haritchabalet (2007) suggested in theory that syndication improves the screening process of
VC deals and prevents competition among investors after investment opportunities are
disclosed. Using a sample of Canadian companies, Brander et al. (2002) provided evidence
that syndicated VC investments have higher returns. Our theory focuses on the endogenous
Page 22 of 48
formation of investment syndicates in equilibrium. It is unique in that it determines the lead
investor’s network capital endogenously, implying a testable formula for network capital.
We focus mainly on the determinants of network capital or syndicate size, which is the
key in our theoretical model. As shown in equation (6), the determinants of the syndicate
size or the lead VC’s network capital ( n ) include the lead VC’s risk aversion ( a ), productivity
(d ) and the lead VC type ( r ). Based on this, we formulate the following regression model:
0 1 2 3 .Dependent Variable Investment Risk Portfolio Company Quality Controlsa a a a= + + +
Here, the dependent variable is the syndicate size or the lead VC’s network capital. Our
theoretical model indicates that the lead VC’s network capital is endogenously determined in
the syndication process. In other words the EN chooses the optimal syndicate size as well as
the lead VC with a certain network capital which is in fact a tradeoff between future funding
and profit allocation. We take the ex post “Investment Risk” as a proxy for the lead VC’s risk
attitude. The higher the lead VC’s risk aversion, the lower the ex post investment risk. “Port-
folio Company Quality” is taken to be a proxy for productivity. The higher the portfolio com-
pany quality, the greater the productivity. “Controls” include lead VC type plus a set of other
control variables.
Equation (6) implies a negative relationship between the syndicate size and the lead
VC’s risk aversion, and a positive relationship between the syndicate size and the firm’s
productivity. Hence, we expect both investment risk and portfolio company quality to be
positively related to the syndicate size.
4.1. Data and Variables
Our sample was obtained from the SDC VentureXpert database, which is the main pub-
lic database for academic research on VC. We collected data on all the VC investments by U.S.
VCs in U.S. portfolio companies that received their initial VC funding during the period
1985–2005. We focused on VC investments that were made in the development stage of a
portfolio company. Therefore, we excluded those investments made in the mature stages,
including investments in “Buyouts/Acquisitions,” “unknowns” and those made by private
equity, such as angel or buyout funds. We also excluded those observations without regres-
sion variables. We eventually identified 40,395 VC investment rounds in 15,264 portfolio
companies.
Table 1 presents the distribution of our sample by funding year and the industry, loca-
tion and development stage of the portfolio company. The number of observations and the
corresponding percentages are also given. As observed, VC investments increased rapidly
Page 23 of 48
since 1994, peaked in year 2000, and slowed down considerably following the “bubble burst”
of the dot-com era. The sample covered the following 18 industries based on the venture
economics industry classification (VEIC) code in the VentureXpert database: agricul-
ture/forestry/fishing, biotechnology, business services, communications, computer hard-
ware, computer other, computer software, construction, consumer related, financial services,
industrial/energy, internet specific, manufacturing, medical/health, semiconductor
/electronics, transportation, utilities, and other. Clearly, VC investments were concentrated
in high-technology industries. In particular the computer software, internet specific, medi-
cal/health, and communications industries constituted respectively 23.0%, 18.3%, 12.9% and
11.3% of the whole sample.
Regarding the locations of the portfolio companies, we can see a clear trend of geo-
graphic clustering in VC investments. For this reason, we report only the top 20 states in
which VC investment rounds have been made. The other 30 states constituted only 7.2% of
the sample. We found that most VC investments occurred in California, Massachusetts,
Texas and New York, which constituted respectively 36.8%, 11.4%, 5.8% and 4.5% of the
sample. That’s a total of almost 60% of our sample. Also, about 40% of the VC investment
rounds were made in the seed or early development stage of a portfolio company, which is
comparable with the 39% reported by Gompers (1995).
The definitions and measures of all the variables are defined and explained in Table 2.
The variables are defined in the remainder of this section.
Dependent Variables
Our unit of analysis is a VC investment round. The dependent variable is the logarithm
of the number of VCs participating in an investment round and is named Syndicate Size. For
each VC investment round, we identified the lead VC as the VC firm that injected the largest
funding. In practice, the lead VC is typically the most active investor and plays a leading role
in monitoring and professionalizing a portfolio company. Our definition of the lead VC is
similar to those of SØrensen (2007), Hochberg et al. (2007) and Nahata (2008). In total,
there were 2,373 lead VCs in our sample.
The network capital in our theoretical model can be in many forms, including other VCs,
head hunters, patent lawyers and investment bankers. In the spirit of Hochberg et al. (2007),
we focused on the most important coinvestment network. We measured a lead VC’s network
capital in a given investment round by the number of VCs that she had syndicated with
during the five years prior to the investment round, normalized it by the number of active
VCs participating in at least one investment round during the five-year span and named it
Page 24 of 48
Network Degree. Network Degree is an indirect measure of a lead VC’s network capital. The
more coinvestment ties a lead VC has, the more opportunities exist for exchange and so the
more influential the lead VC. A lead VC who has many coinvestment ties with other VCs may
have access to a wide range of experts, contacts and pools of capital. To further analyze a lead
VC’s direct relationships with other VCs, we constructed two alternative variables indicating
the lead VC’s network capital in a given investment round. Specifically Network Outdegree
and Network Indegree are measured respectively as the normalized number of VCs a lead
VC had invited into her syndicates and the number of VCs who had invited the lead VC into
their syndicates in the five years prior to the investment round. These three measures repre-
senting the lead VC’s network capital are the same as those used by Hochberg et al. (2007).
For example, if a lead VC had syndicated with 10 different VCs, had invited five different VCs
into her syndicates and had been invited into four different VCs’ syndicates during 2000–
2004, and if there were a total of 21 active VCs participating in at least one investment deal
during 2000–2004, the values of Network Degree, Network Outdegree and Network Inde-
gree of the lead VC for those investment rounds occurring in 2005 would respectively be
50% (10/(21-1)), 25% (5/(21-1)) and 20% (4/(21-1)).
Investment Risk
Investment risk was used to represent a lead VC’s risk version ( a ). The first measure of
investment risk in an investment round, named Company Age, is the logarithm of the age of
a portfolio company at the time of the investment round, while the age of a portfolio com-
pany at the time of the investment round is the number of years between its founding and
the year of the investment round. Hence, a younger company is likely to be riskier.
It is widely observed that VCs typically target certain industries, certain development
stages of the portfolio companies, or certain local geographical areas (Bygrave, 1987; Gupta
and Sapienza, 1992; Norton and Tenenbaum, 1993). This may help them control risk when
identifying, evaluating and monitoring projects and prevent competition from other VCs and
other types of financial institutions. Our sample distribution in Table 1 also implies such
investment clustering. In the spirit of Sorenson and Stuart (2001), investment risk was de-
termined by industry distance, stage distance and geographic distance between a lead VC
and her portfolio company. A lead VC who makes an investment in a faraway portfolio com-
pany, a new industry, or a new development stage of a portfolio company is bearing more
investment risk than most and so she must be less risk averse.
We measured the Industry Distance of a given investment round by the percentage of
investment rounds a lead VC had made from 1980 to the year prior to the given investment
Page 25 of 48
round that were not in the same industry as the portfolio company. There are totally 18
industry groups based on the VEIC code. This measure incorporates at least five years of VC
investment experience. For a VC investment round that occurred in 2000, this measure of
industry distance incorporates 20 years of VC investment experience during 1980–1999. For
example, if the SDC VentureXpert database shows that a lead VC had made a total of 100
rounds of investment during 1980–1999 and 15 of those were in the “Computer Hardware”
industry, the Industry Distance of this lead VC for an investment round made in 2000 and
in a portfolio company in the “Computer Hardware” industry would be 85% ((100-15)/100).
Similar to Industry Distance, we measured the Stage Distance of a given investment
round by the percentage of investment rounds a lead VC had made from 1980 to the year
prior to the given investment round that were not in the same development stage as the
portfolio company in the given round. We covered four development stages: seed stage, early
stage, expansion stage and later stage. As a robustness check, we also measured Industry
Distance and Stage Distance by past investment amount rather than by the number of past
investment rounds. All the results are similar. Finally, we measured Geographic Distance
directly by the logarithm of the physical distance between the state in which the lead VC is
located and her portfolio company.
Portfolio Company Quality
We took the quality of a portfolio company as a proxy for project productivity ( d ) in the
theoretical model. The greater the total VC investment, the higher the portfolio company
quality is likely to be, and the higher the likelihood of good performance. Gompers (1995)
proposed that funding in follow-on rounds is given only if a portfolio company does well in
earlier rounds. Also, specific milestone-contingent clauses are contained in most VC financ-
ing contracts, by which VCs automatically cease to provide further funding to weak ventures
(Wang, 2009). Hence, a larger VC funding reflects better quality of the portfolio company.
Following Nahata (2008), we used the logarithm of the total VC funding amount in a portfo-
lio company across all financing rounds as a proxy for its quality and named it Total VC
Funding.
We also defined an indicator variable on whether a portfolio company has successfully
exited through an IPO or acquisition as a proxy for its quality and named it Quality Indica-
tor. In practice, observable VC returns are mainly made from successful exits, such as if the
portfolio company goes public or is acquired (Cumming and MacIntosh, 2003; Cochrane,
2005). The larger the number of successful exits a VC makes from its portfolio companies,
the larger its internal rate of return from investments. Hence, we also used Quality Indicator
Page 26 of 48
to represent portfolio company quality. We traced the exits of portfolio companies until the
end of 2009, which provides a minimum of four years for a successful exit.
These two measures of company quality are by no means perfect and involve a look-
ahead bias since the cumulative funding and a successful exit are unknown at the time of
early financing rounds. However, considering that VentureXpert provides little information
on portfolio companies other than financing rounds, funding amount, investors and exits,
these two measures have been well adopted in the literature and have been considered to
reflect many unobservable factors that determine a portfolio company’s prospects at the time
of funding.
Control Variables
Other than the explanatory variables indicating investment risk and portfolio company
quality, we also needed to control for lead VC type based on its organizational structure.
There is a distinction between traditional VCs and the so-called captive VCs. The latter are
affiliated with corporations, financial institutions or governments (Bottazzi et al., 2008;
Nahata, 2008; Hellmann et al., 2008). Different types of VCs may have different resources
and connections with other institutions and emphasize different strategic objectives. To
control for lead VC type, we created three dummy variables indicating whether a lead VC was
affiliated with a corporation, an institution, or a government, and named them Corporate VC
Indicator, Institutional VC Indicator and Government VC Indicator, respectively.
We also included industry indicators based on the VEIC code to partially account for
technological and industrial characteristics of portfolio companies. We further included
funding year indicators as additional control variables. Considering the possible effect of an
early choice of syndication on later rounds, we further controlled for the round sequence
effects based on the sequencing of investment rounds in each portfolio company (being in
the first, second or third round, etc.).
4.2. Summary Statistics and Correlations
Table 3 presents summary statistics of all the regression variables.7 The quartiles, means,
standard deviations and the number of observations are presented. As indicated, about 2.4
VCs formed a syndicate in an investment round on average, suggesting the popularity of
7 We reported raw values for all the variables in Table 3, even though some of them were used in their natu-
ral logarithm form in our regressions (as indicated by “**” in Table 2).
Page 27 of 48
syndication in VC investment. About 54% of VC investment rounds were syndicated (not
tabulated). A typical lead VC syndicated with 7.48% of all active VCs (Network Degree) on
average, invited 4.71% of all active VCs to her syndicates (Network Outdegree), and was
invited by 1.37% of all active VCs to their syndicates (Network Indegree) in the five years
prior to the current investment round. The lead VCs in our sample had higher network capi-
tal on average than those reported by Hochberg et al. (2007). In particular, Hochberg et al.
(2007) reported the averages of Network Degree, Network Outdegree and Network Inde-
gree to be 4.24%, 1.20% and 1.00%, respectively. The reason for the difference is that we
measured the network capital for the lead VCs only, who tend to have more network capital
and their network capital tends to be more influential than an average VC.
Table 3 indicates high investment risk in the VC market. On average, a portfolio com-
pany was less than five years old with the median being only three years old. A trend in
specialization is clear. In an investment round, on average 20% (or 32.5%) of a lead VC’s
investment experience was in the industry (or in a development stage) of the portfolio com-
pany. In other words, the average Industry Distance was 80% (with a standard deviation of
24%), which is very close to the 77% (with a standard deviation of 21%) reported by Sorenson
and Stuart (2001). The distribution of Stage Distance was similar to that of Industry Dis-
tance, whose mean and standard deviation were 67.54% and 21%, respectively. The average
geographic distance between a lead VC and her portfolio company was about 1,208 kilome-
ters with the median being only 363.9 kilometers. Also, 41% of VC investment rounds were
made in the same state as the lead VC’s (unreported). For portfolio companies, the total VC
fund inflows across all rounds were 22.27 million dollars on average,8 which is comparable
with the 24 million dollars reported by Nahata (2008). As observed, 38% of VC investments
made successful exits through IPOs or acquisitions by the end of 2009, which is slightly
higher than the 35% reported by Gompers (1995) for the period 1961–1992. Finally, the
probabilities that a lead VC was affiliated with a corporation, a financial institution and a
government were respectively 3%, 4% and 11%.
The Spearman and Pearson correlation matrices are tabulated respectively in the upper
and lower diagonals of Table 4. The correlations among all the variables were small, except
for those among the three alternative measures of the network capital of the lead VC, namely
Network Degree, Network Outdegree and Network Indegree, which suggests the absence of
severe multicollinearity in our regression model.
8 The value of Total VC Funding reported in Table 3 was based on VC investment rounds, which is slightly
different from the value based on portfolio companies.
Page 28 of 48
4.3. Regression Results
We also investigated the impact of investment risk and portfolio company quality on the
syndicate size (which also indicates the network capital of a lead VC in equilibrium in our
theoretical model). Our unit of analysis was a VC investment round. The sample size was
40,395. We took Syndicate Size and the three alternative measures of the network capital of
the lead VC, namely Network Degree, Network Outdegree and Network Indegree, as the
dependent variables. We employed an OLS model to relate the dependent variables with the
following primary explanatory variables: investment risk as measured/proxied by Company
Age, Industry Distance, Stage Distance and Geographic Distance, and portfolio company
quality as measured/proxied by Total VC Funding and Quality Indicator. We also controlled
for lead VC type by including Corporate VC Indicator, Institutional VC Indicator and Gov-
ernment VC Indicator in all the regressions. Although not reported, in all the regressions, we
also included 20 funding year indicators, 17 industry indicators based on the VEIC codes and
19 round sequence indicators based on the sequencing of investment rounds in each portfo-
lio company (being in the first, second or third round, etc.) to control for the year, industry
and round sequence fixed effects.
Table 5 reports the results. All the t-statistics have been adjusted for heteroskedasticity
(White, 1980). In models 1 and 2, we took Syndicate Size as the dependent variable and
separately included factors representing investment risk and portfolio company quality. In
models 3 to 5, we separately took the three alternative measures of the network capital of a
lead VC as the dependent variable and included all the explanatory variables. We found very
similar results in all the models. Our empirical results are consistent with our theoretical
findings.
Our theoretical model implies a negative relationship between a lead VC’s risk aversion
( a ) and the syndicate size ( n ). Our regression results confirm this prediction by showing
that investment risk had a positive impact on the syndicate size. Also, the coefficient of Com-
pany Age was negative at the 1% significance level and the coefficients of Industry Distance,
Stage Distance and Geographic Distance were all significantly positive at the less than 5%
significance level. These results are also consistent with the prediction of our theoretical
model.
For the magnitude of the effects, we found that a one standard deviation increase in
Company Age decreases Syndicate Size and its three alternative measures, Network Degree,
Network Outdegree and Network Indegree, by 1.83%, 48.31%, 22.82% and 8.97%, respec-
tively. Also, a one standard deviation increase in Industry Distance (Stage Distance, Geo-
graphic Distance) increases Syndicate Size and its three alternative measures, Network
Page 29 of 48
Degree, Network Outdegree and Network Indegree, by 3.29% (1.40%, 2.11%), 108.81%
(12.10%, 37.13%), and 18.18% (3.08%, 1.49%), respectively.
Our theoretical model implies that productivity ( d ) has a positive impact on a lead VC’s
equilibrium network capital ( n ). Hence, we expect a positive relationship between portfolio
company quality and the syndicate size in our regressions. Indeed, the coefficients of Total
VC Funding and Quality Indicator were both positive at the 1% significance level. Specifi-
cally, a one standard deviation increase in the total VC funding across all rounds for a portfo-
lio company increases Syndicate Size and its three alternative measures, Network Degree,
Network Outdegree and Network Indegree, by 19.91%, 108.42%, 64.07% and 15.02%, re-
spectively. Also, a one standard deviation increase in the likelihood of a successful exit im-
plies an increase in Syndicate Size and its three alternative measures, Network Degree,
Network Outdegree and Network Indegree, of 6.05%, 32.83%, 22.46% and 4.08%, respec-
tively.
Since lead VC type ( r ) in our theoretical model can be measured by a set of characteris-
tics of the lead VC in our empirical model, our theory predicts an impact, but not the specific
signs of the coefficients of those variables representing the lead VC’s characteristics. Indeed,
as expected, our regressions indicate that both Corporate VC Indicator and Government VC
Indicator had a negative impact, but Institutional VC Indicator had a positive impact, on
Syndicate Size.
In summary, after controlling for lead VC type and the fixed effects of funding year, in-
dustry and round sequence, we found that both investment risk and portfolio company
quality had a significant positive effect on syndicate size. This empirical finding is consistent
with our theoretical finding.
4.4. Robustness Checks
To ensure the reliability of our empirical analysis, we conducted many robustness
checks. Firstly, we investigated the impact of investment risk and portfolio company quality
on the syndicate size at the portfolio company level to rule out the double counting problem.
The unit of analysis became a portfolio company, rather than a VC investment round. We
defined Syndicate Size as the number of VCs providing funding to a portfolio company. The
lead VC was identified as the VC that participated in the initial round of financing and made
the largest total investment in the portfolio company across all rounds. The lead VC’s net-
work capital was measured similarly as before but this time we took its value at the time a
portfolio company received initial VC funding. With a reduced sample size of 15,264, all our
results remained qualitatively the same after these adjustments.
Page 30 of 48
Secondly, we also tried many alternative measures of investment risk and portfolio
company quality. For example, we measured Industry Distance and Stage Distance based on
the past investment amount rather than the number of past investment rounds. We rede-
fined Geographic Distance as a dummy variable indicating whether a lead VC and her port-
folio company were located in the same state. We further included public companies’ risk as
a proxy for the investment risk borne by these private companies. Specifically, we first iden-
tified the 4-digit SIC code for each portfolio company based on the VEIC-SIC concordance
provided by Dushnitsky and Shaver (2009). Then we calculated the standard deviation of the
ROA, ROE and stock returns of public companies using accounting and stock information
from Compustat and CRSP with the same 2-digit SIC code of the portfolio company in the
year prior to a VC investment round. Lastly we took these industry risk variables of the pub-
lic companies as proxies for the investment risk of the VC investment round. All our findings
remained unaltered.
On portfolio company quality, around 25% of the portfolio companies voluntarily re-
ported their market valuation after an investment round, which is a good predictor of their
quality. Even though we do not know exactly the reasons for disclosing company market
valuation, we argue that portfolio companies with better quality are more likely to disclose
their market valuation. It is widely suggested in the accounting literature that voluntary
information disclosure plays a role in mitigating asymmetric information and agency prob-
lems (Myers and Majluf, 1984). This reduces the cost of capital (Botosan, 1997) and IPO
underpricing (Leone, et al., 2007) and increases market capitalization of earnings (Piotroski,
1999). Therefore, we used a dummy variable indicating whether a portfolio company dis-
closes its post-round market valuation as one more proxy for company quality. We found
that this variable had a positive effect on Syndicate Size at the 1% significance level. Again,
all our existing results remained unchanged.
Thirdly, to investigate further whether our results were unduly influenced by outliers,
we re-ran our regressions after winsorizing the top and bottom 1%, 2% and 5% for each
variable. The results remained largely similar.
Fourthly, considering the possible impact of the NASDAQ bubble during 1999–2000, we
re-ran our regressions by excluding these two years. The sample size was reduced to 31,368.
Our results were qualitatively unchanged.
Finally, as shown in Table 1, the computer software industry and the California state
made the largest contributions to our dataset respectively in terms of the portfolio company
industry and location. To examine whether the inclusion of these special subsamples influ-
enced our results, we ran our regressions yet again after excluding the software industry or
Page 31 of 48
the California state, which reduced the sample size respectively to 31,115 and 27,731. The
results again remained qualitatively unchanged.
5. Conclusion
As an effective investment strategy, investors often form a syndicate to invest jointly in a
company. The unique feature of this paper is that it endogenizes the formation of the in-
vestment syndicate. We provide a theory on endogenous networks in investment syndication
and analyze how several key factors such as risk aversion, productivity, risk and cost affect
the incentives for and the investments by managers and investors. Since investment syndica-
tion is very popular in VC investments, we also apply our theory to VC investments and
identify empirical evidence in support of our theory.
Unlike previous empirical papers on investment syndication or investment networks
which have mostly focused on the role of the lead investor (such as Brander et al., 2002;
Hochberg et al., 2007, 2010; Casamatta and Haritchabalet, 2007), this paper focused instead
on strategic interactions between managers and investors. In our model, investment deci-
sions, together with syndication and network, are determined in equilibrium in three stages.
Most of the existing studies have focused only on the second stage. In addition, we demon-
strated that network endogeneity is crucial. For example, productivity can have opposite
effects on investment if the network is treated as exogenous instead of endogenous (Figures 1
and 2).
We have also applied our theory to VC investments. Using 40,395 domestic VC invest-
ment rounds made in 15,264 U.S. portfolio companies by U.S. VCs during 1985–2005 and
controlling for various factors, we found empirical support for our theory.
Syndication occurs in many domains, including the VC industry, IPOs, bond issuing, re-
al estate development, and large infrastructure projects (subways, railways, roads and air-
ports). Although our empirical work focuses on VC, insights from our study provide an un-
derstanding of syndication as a form of network in general.
Appendix
In this appendix, we give the derivations for the expressions in (3), (6)–(8) and (9).
Page 32 of 48
A.1. Derivation of (3)
The FOC for 1I is
( )
( )
11 11 2
22
1 ( ) ( ) ( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( ) ( ) ( ) .
jj
jjjj
ni i i
ii jjjj
I II Ip a u n x I n x I n x II II
I I Ip a u n x I n x I n x II II
l l l
l l l=
ì üï ï-é ù ï ïï ï¢ ¢ê ú= +í ýê úë û ï ïï ïï ïî þé ù ì üï ïï ï¢ ¢ê ú+ - +ï ïí ýê úë û ï ïï ïï ïî þ
ååå
å åå
With the same risk aversion a for all the investors, we must have iI nI= for all i in equilib-
rium. Then,
{ }
{ }
1 2
22
1
11 11 ( ) ( ) ( ) ( ) ( ) ( ) ( )
1 1( ) ( ) ( ) ( ) ( ) ( ) ( )
1 1 1( ) ( ) ( ) ( ) ( ) ( ) ( )
1( ) ( ) ( )
n
ii
i
I Inp a u n x I n x I n x I
n I nIp a u n x I n x I n x I
n nI nnp a u n x I n x I n x I
n nI n
p a u n x In
l l l
l l l
l l l
l
=
ì üï ïï ï-ï ïé ù í ý¢ ¢ê ú= +ï ïï ïê ú ï ïë û î þé ù
¢ ¢ê ú+ - +ê úë û
é ù -¢ ¢ê ú= +ê úë û
é¢ ê+ë
å
{ }2
1 1( ) ( ) ( ) ( ) ,n
in x I n x I
nI nl l
=
ù¢ú - +
ê úûå
implying
[ ]11
1 1( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) .n
ii
nI n n p a x I u n x I p a u n x I I n x I n x In n
l l l l l=
é ù é ù¢ ¢ ¢ê ú ê ú= + -ê ú ê úë û ë û
å
With the same utility function for all, we have
[ ]1 1( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ,nI n n p a x I n x I p a n n x I I n x I n x In n
a a
l l l l l- -é ù é ù
¢ê ú ê ú= + -ê ú ê úë û ë û
implying
1 1( ) ( ) ( ) ( ) ( ) 1 ,
( ) ( )I n x I n x I Ix I
p a n n n x I
a al l- - é ù¢é ù é ùê úê ú ê ú= + -ê úê ú ê úë û ë û ë û
which implies (3).
A.2. Derivation of the Solution (6)–(8)
Condition 2IC implies
Page 33 of 48
1
1( ) .n a n In
ag d adl
d-
- +é ùê ú =ê úë û
(23)
By this, condition 1IC implies
[ ]1 1( )1 ( ) ,nn a n a
n
da d ad
g b gll g d b
- - +-
é ùæ öê ú÷ç- =÷ç ÷çê úè øë û
implying
[ ]
1( )(1 )
1 1( )1 ( ) .na n nn
d add b g ad bd
a d adg ll d
b
- +- + -
- - +ì üï ïï ïé ùï ïæ öí ýê ú÷ç= - ÷ï ïç ÷çê úè øï ïë ûï ïî þ
(24)
By (23) and (24), we find
[ ]
( )
[ ]( )
1( )(1 )
1 1 11
1(1 ) 1( )(1 ) ( )(1
( ) ( )1 ( )
( ) 1 ( )
n nI n n nn n
n nn
d ad gd b g ad bd
a a d add ad
gd d ad ga
b g ad bd b g ad
l g ld l d
b
l gl
b
- +- + -
- - - +- +
æ ö - +÷ç- + ÷ç ÷ç ÷ç - + - - +è ø
ì üï ïï ïé ùï ïæ ö æ öí ýê ú÷ ÷ç ç= -÷ ÷ï ïç ç÷ ÷ç çê úè ø è øï ïë ûï ïî þ
æ ö ì üï ïï ï÷ç= -÷ í ýç ÷çè ø ï ïï ïî þ( )
) 1( )(1 ) .n
gdbdb g ad bdd
- +- + -
(25)
Then, the objective function in problem (4) becomes
[ ]
( ) ( )
( )
1 1 (1 )( )(1 ) ( )(1 ) ( )(1 )
(1 )( )(1 ) ( )(1 )
1 ( )
[1 ( )] .
n a I an
n n
ad g d ad b a bdb g ad bd b g ad bd b g ad bdg d b
b d ad bdb g ad bd b g ad bd
g g ll
b b
l d
+ - + -- + - - + - - + -
- +- + - - + -
é ùæ ö æ ö æ öê ú÷ ÷ ÷ç ç ç- - = -ê ú÷ ÷ ÷ç ç ç ÷ç÷ ÷ç ç è øê úè ø è øë û
⋅ -
(26)
Since ,g b< we have
( ) ( )1 1( )(1 ) ( )(1 )
,ad g d ad b
b g ad bd b g ad bdg gb b
+ - +- + - - + -æ ö æ ö÷ ÷ç ç>÷ ÷ç ç÷ ÷ç çè ø è ø
which guarantees a positive payoff for the EN. Then, problem (4) becomes
( )
(1 ) (1 )( )(1 ) ( )(1 ) ( )(1 )max [1 ( )] .
nn n
n
a bd b d ad bdb g ad bd b g ad bd b g ad bdl
l d
- - +- + - - + - - + -
æ ö÷ç -÷ç ÷çè ø
With ( ) ,n nl r= the above problem becomes
(1 )
( )(1 ) ( )(1 )max (1 ) .n
n nb d ad bd
b g ad bd b g ad bdr- +
- + - - + --
The first-order condition (FOC) then implies
Page 34 of 48
.(1 )
n dad r
* =+
The second-order condition (SOC) is satisfied. Hence, we are sure that the solution in (6) is
correct. Then from (24), we find
( ) ( )1
( )(1 )1 1
1( )(1 )
2 1
12 ( )(1 )( )(1 )
1
11 1
11 1
a n n
d add b g ad bd
a d ad
d add b g ad bd
d ada
dd adb g ad bdb g ad bd a
gr r d
b
ad d g dr
ad b ad
ad d g dr
ad b ad
- +- + -
* * - * - +
- +- + -
- +-
- +- + -- + - -
é ùê ú= -ê úë û
é ùæ öê ú+ - ÷çê ú÷= ç ÷÷çê úè ø+ +ë û
æ öæ ö+ - ÷ç÷ç ÷= ÷ çç ÷÷÷ çç è ø+ +è ø.
By (25), we have
( )
( )
1 12(1 ) 1 ( )(1 )( )(1 )( )(1 )1
1 12 ( )(1( )(1 )
11 (1 )
11 1
I
gdd ad ggda b g ad bdb g ad bdb g ad bdd ad
gdd ad gb gb g ad bd a
ad d g dr
ad b ad r
ad d g dr
ad b ad
- +æ ö +÷ç- + ÷ - + -ç - + -÷ç ÷ç - + -- + è ø
- + +-- + - -
æ öæ ö+ - ÷ç÷ç ÷= ÷ çç ÷÷ çç ÷+ +è ø è ø
æ öæ ö+ - ÷ç÷ç ÷= ÷ çç ÷÷÷ çç è ø+ +è ø
),
ad bd+ -
implying
2 ( )(1 )( )(1 )1 .
1 1I
b ggb g ad bdb g ad bd aad d g d
rad b ad
-- + -- + -* -
æ öæ ö+ - ÷ç÷ç ÷= ÷ çç ÷÷÷ çç è ø+ +è ø
A.3. Derivation of (9)
With a fixed ,n (24) and (25) imply
( ) ( )
( )( )
( )
1( )(1 )
1 1
1(1 ) 1 ( )(1 ) 1( )(1 )1 ( )(1 )
ˆ 1 ,
ˆ 1 ,
a n n
I n n
d add b g ad bd
a d ad
d ad ggd gda b g ad bdb g ad bdd ad b g ad bd
gr r d
b
gr r d
b
- +- + -
- - +
- +æ ö÷ç- + ÷ç - + -÷ç +÷ç - + -- + è ø - + -
ì üï ïï ï= -í ýï ïï ïî þ
ì üï ïï ï= -í ýï ïï ïî þ
which in turn imply the solution in (9).
A.4. Derivation of (10)
The FOC is
Page 35 of 48
( ) ( )( )
( ) ( )2
( ) ( ) ( )1 ( ) ,( )
i i ii j j j jj j j j
j j jj j jjj
n I n I n Inu x I x I x I x Ip a I I II
l l llé ù é ù¢ ¢ê ú ê ú= - +ê ú ê úê ú ê úë û ë û
å å å åå å åå
or
2
( )1 1( ) ( ) ( ) ( ) ( ) .( )
i i ii
n I I In u x I x I x I x Ip a I I I I
ll
é ù é ù¢ ¢ê ú ê ú= - +ê ú ê úë û ë û
Since we must have /iI I n= for all i in equilibrium, this FOC becomes
1 ( ) ( ) 1 ( )( ) ( ) ,( )
n x I n x In x Ip a n nI n
all
- é ùé ù ¢-ê úê ú= +ê úê úë û ë û
which implies (10).
A.5. Derivation of the Nash Investment Solution
Condition 2IC implies
( )1 11 ,na II I
ag d dr
- æ ö- ÷ç= + ÷ç ÷çè ø
implying
( )1 1 1 .I a nd ad g ar d- + -= + - (27)
By this, condition 1IC implies
( )1 1(1 ) 1 ,n a n ad
g a b gd adr g r d b- -- +é ù- + - =ë û
implying
( )1 11(1 ) 1 ,n n agdd b g
a d add adg
r r db
- -- - +- +é ù- + - =ë û
implying
( )
1( )(1 )
1 1ˆ (1 ) 1 .a n n
d add b g ad bd
a d adg
r r db
- +- + -
- - +ì üï ïï ïé ù= - + -í ýë ûï ïï ïî þ
(28)
By (27) and (28),
Page 36 of 48
( ) ( )
( )
( )
( )
1( )(1 )
1 1 1 1
1( )(1 )11 ( )(1 )
1 (1 ) 1
1 (1 ) .
I n n n
n n
d ad gd b g ad bd
d ad a a d ad
d ad gdg b g ad bda b g ad bd
gr d r r d
b
gr d r
b
- +- + -
- + - - - +
- +- + -+- - + -
ì üï ïï ïé ù= + - - + -í ýë ûï ïï ïî þ
é ùé ù ê ú= + - -ë û ê úë û
(29)
Then,
( )( )(1 )1 ( )(1 )ˆ 1 (1 ) .I n n
gb g b g ad bda b g ad bd
gr d r
b
- - + -- - + -é ùé ù ê ú= + - -ë û ê úë û
Then, the objective function in problem (11) becomes
( )
( )
( )
( )
( )
1( )(1 )
1 1
( ) ( )(1 )1 ( )(1 )
1(
1 1
(1 ) (1 ) (1 ) 1
1 (1 )
(1 ) 1
n a I a n n n
n n
n n
d ad gd b g ad bd
g d b a d ad
gdb g d b g ad bda b g ad bd
d ad bd
a d ad
gr r r r d
b
gr d r
b
gr r d
b
- +- + -
- - +
- - + -- - + -
- +
- - +
ì üï ïï ïé ù- - = - - + -í ýë ûï ïï ïî þ
é ùé ù ê ú⋅ + - -ë û ê úë û
ì üï ïï ïé ù- - + -í ýë ûï ïï ïî þ( ) ( )
( )( )
( )
)(1 )
1 11( )(1 ) ( )(1 ) 1 ( )(1 )( )(1 )
,
1 1 .n n
b g ad bd
ad g d ad bbdd ad bb g ad bd b g ad bd a b g ad bdb g ad bd
g gr r d
b b
- + -
+ - +- +- + - - + - - - + -- + -
é ùæ ö æ öê ú÷ ÷ é ùç ç= - - + -ê ú÷ ÷ç ç ë û÷ ÷ç çê úè ø è øë û
(30)
Since ,g b< we have
( ) ( )1 1( )(1 ) ( )(1 )
,ad g d ad b
b g ad bd b g ad bdg gb b
+ - +- + - - + -æ ö æ ö÷ ÷ç ç>÷ ÷ç ç÷ ÷÷ ÷ç çè ø è ø
which guarantees a positive payoff for the EN. Then, problem (11) becomes
( )( )
( )1
( )(1 ) ( )(1 )max 1 1 .n
n nd ad b bd
b g ad bd b g ad bdr d- +
- + - - + -- + -
The first-order condition (FOC) is
1 ,
1 1n nd d ad
rd r
- +=
+ - -
implying
(1 )(1 ) .
(1 )n d r d ad d
ad r* + - + -
=+
The second-order condition (SOC) is satisfied. Hence, we are sure that this solution is correct.
Page 37 of 48
A.6. Proof of Proposition 1
Part (a)
We have
( )( )1 0 1 .1
g db g ad bd
b ad- + - > - >
+ (31)
If 1 ,1
g db ad
- >+
then
( ) ( )2
11ˆ(1) 1 ,1 1
I Ib gg g
b ga aad d g d gr r dr
ad b ad b
--* - -
æ öæ ö é ù+ - ÷ç÷ç ÷ ê ú³ ³ -÷ çç ÷÷÷ çç è ø ê ú+ +è ø ë û
or
( ) ( )1 1 1 .b g
bgg
rr ad ad d
d
-
æ ö÷ç- + £ + -÷ç ÷çè ø (32)
Hence, if 1 ,1
g db ad
- >+
then
( ) ( )ˆ(1) 1 1 1 ;I Ib g
bgg
rr ad ad d
d
-
* æ ö÷ç³ - + £ + -÷ç ÷çè ø
and if 1 ,1
g db ad
- <+
then
( ) ( )ˆ(1) 1 1 1 .I Ib g
bgg
rr ad ad d
d
-
* æ ö÷ç£ - + £ + -÷ç ÷çè ø (33)
We also have
2
(1 ) (1 ) 0 1 .b g b b g
g g gb g gr r r r r r
r g b
- --é ù¶ -ê ú- = - - ³ £ -ë û¶
Therefore, if 1 ,gr
b£ - then
1d
rad
£+
implies
( ) ( ) ( )11 1 1 1 1 .1 1
b g b gb bg gg g
r dr ad ad ad d
d ad ad
- -
æ ö æ öæ ö÷ ÷ ÷ç ç ç- + £ - + = + -÷ ÷ ÷ç ç ç÷ç ÷ ÷ç çè ø è øè ø+ +
Page 38 of 48
That is, if 1 gr
b£ - and 1 ,
1g db ad
- >+
given the c0ndition 1
dr
ad£
+ (i.e., 1),n* ³ we
have ˆ(1)I I* ³ . But, if 1 gr
b£ - and 1 ,
1g db ad
- £+
given the c0ndition ,1
dr
ad£
+ we
have ˆ(1).I I* £
On the other hand, if 1 ,gr
b³ - then
1d
rad
£+
implies
( ) ( ) ( )11 1 1 1 1 .1 1
b g b gb bg gg g
r dr ad ad ad d
d ad ad
- -
æ ö æ öæ ö÷ ÷ ÷ç ç ç- + ³ - + = + -÷ ÷ ÷ç ç ç÷ç ÷ ÷ç çè ø è øè ø+ + (34)
Given ,1
dr
ad£
+ if 1 ,g
rb
³ - then we have 1 .1
g db ad
- £+
By (33) and (34), we then
have ˆ(1).I I* ³ Part (a) is proven.
Part (b)
We have
( )( )1 0 1 .1
g db g ad bd
b ad- + - > - >
+ (35)
If 1 ,1
g db ad
- >+
then
( ) ( )1 12
11ˆ(1) 1 ,1 1
a add ad d ad
da aad d g d gr r dr
ad b ad b
- + - +* - -
æ öæ ö é ù+ - ÷ç÷ç ÷ ê ú> > -÷ çç ÷÷÷ çç è ø ê ú+ +è ø ë û
or
( ) ( )11
11 1 1 .d
add add ad
rr ad ad d
d
+- +- +
æ ö÷ç- + < + -÷ç ÷çè ø
Hence, if 1 ,1
g db ad
- >+
then
( ) ( )11
1ˆ(1) 1 1 1 ;a ad
add add ad
rr ad ad d
d
+- +* - +æ ö÷ç³ - + £ + -÷ç ÷çè ø
(36)
and if 1 ,1
g db ad
- <+
then
Page 39 of 48
( ) ( )11
1ˆ(1) 1 1 1 .a ad
add add ad
rr ad ad d
d
+- +* - +æ ö÷ç³ - + ³ + -÷ç ÷çè ø
(37)
We also have
1
1 1 1(1 ) (1 ) 0 .1 1
d d dd ad d ad d add d
r r r r r rr d ad ad
-- + - + - +
é ù¶ ê ú- = - - ³ £ë û¶ - + +
The given condition is 1n* ³ or ,1
dr
ad£
+ which ensures that 1(1 )
dd adr r - +- is increasing
in .r Then, 1
dr
ad£
+ implies
( ) ( ) ( )1 11 1
1 111 1 1 1 1 .
1 1
d dad add ad d ad
d ad d adr d
r ad ad ad dd ad ad
+ +- + - +- + - +
æ ö æ öæ ö÷ ÷ ÷ç ç ç- + £ - + = + -÷ ÷ ÷ç ç ç÷ç ÷ ÷ç çè ø è øè ø+ +
Hence, if 1 ,1
g db ad
- >+
by (36), we have ˆ(1);a a* ³ and if 1 ,1
g db ad
- <+
by (37), we
have ˆ(1).a a* £ Part (b) is proven.
References
Admati, A.R., Pfeiderer, P., 1994. Robust Financial Contracting and the Role of Venture
Capitalist. Journal of Finance 49(2), 371–402.
Afonso, G.M., Shin, H.S., 2007. Systemic Risk and Liquidity in Payment Systems. Work-
ing paper, London School of Economics.
Allen, F., Babus, A., 2009. Networks in Finance. The Network Challenge, edited by
Kleindorfer, P., Wind, J., Wharton School Publishing, 367–382.
Allen, F., Gale, D., 2000. Financial Contagion. Journal of Political Economy 108(1), 1–
33.
Amit, R., Glosten, L., Muller, E., 1990. Entrepreneurial Ability, Venture Investments,
and Risk Sharing. Management Science 36(10), 1232–1245.
Babus, A., 2009. The Formation of Financial Networks. Working paper, University of
Cambridge.
Page 40 of 48
Botosan, C.A., 1997. Disclosure Level and the Cost of Equity Capital. Accounting Review
72(3), 323–350.
Bottazzi, L., Da Rin, M., Hellmann, T., 2008. Who Are the Active Investors? Evidence
from Venture Capital. Journal of Financial Economics 89(3), 488–512.
Brander, J.A., Amit, R., Antweiler, W., 2002. Venture Capital Syndication: Improved
Venture Selection vs. the Value-Added Hypothesis. Journal of Economics and Management
Strategy 11(3), 423–452.
Bygrave, W.D., 1987. Syndicated Investment by Venture Capital Firms: A Networking
Perspective. Journal of Business Venturing 2(2), 139–154.
Casamatta, C., Haritchabalet, C., 2007. Experience, Screening and Syndication in Ven-
ture Capital Investments. Journal of Financial Intermediation 16(3), 368–398.
Castiglionesi, F., Navarro, N., 2008. Optimal Fragile Financial Networks. Working Pa-
per, Tilburg University.
Cestone, G., Lerner, J., White, L., 2006. The Design of Syndicates in Venture Capital.
Working Paper, Harvard Business School.
Cifuentes, R., Ferrucci, G., Shin, H.S., 2005. Liquidity Risk and Contagion. Journal of
European Economic Association 3(2-3), 556–566.
Cochrane, J.H., 2005. The Risk and Return of Venture Capital. Journal of Financial
Economics 75(1), 3–52.
Cumming, D.J, MacIntosh, J.G, 2003. A Cross-Country Comparison of Full and Partial
Venture Capital Exits. Journal of Banking and Finance 27(3), 511–548.
Cumming, D.J, Walz, U., 2010. Private Equity Returns and Disclosure around the World.
Journal of International Business Studies 41(4), 727–754.
Cummins, D.J, Doherty, N., Lo, A., 2002. Can Insurers Pay for the “Big One”? Measur-
ing the Capacity of the Insurance Market to Catastrophic Losses. Journal of Banking and
Finance 26 (2-3), 557–583.
Dasgupta, A., 2004. Financial Contagion through Capital Connections: A Model of the
Origin and Spread of Bank Panics. Journal of European Economic Association 2(6), 1049–
1084.
De Clercq, D., Dimov, D.P., 2004. Explaining Venture Capital Firms’ Syndication Behav-
ior: A Longitudinal Study. Working Paper, Ghent University.
Page 41 of 48
Dorobantu, F.A., 2006. Syndication and Partial Exit in Venture Capital: A Signaling Ap-
proach. Working paper, Duke University.
Du, Q., 2009. Birds of Feather or Celebrating Differences? The Formation and Impact of
Venture Capital Syndication. Working Paper, Shanghai Jiao Tong University.
Durlauf, S.N., 2004. Neighborhood effects. Handbook of Regional and Urban Econom-
ics, edited by Henderson, J. V., Thisse, J. F., Elsevier Press.
Dushnitsky, G., Shaver, J.M., 2009. Limitations to Inter-Organizational Knowledge Ac-
quisition: The Paradox of Corporate Venture Capital. Strategic Management Journal 30(10),
1045–1064.
Eisenberg, L., Noe, T.H., 2001. Systemic Risk in Financial Systems. Management Sci-
ence 47(2), 236–249.
Fluck, Z., Garrison, K., Myers, S., 2006. Venture Capital Contracting: Staged Financing
and Syndication of Later-stage Investments. NBER Working Paper.
Freixas, X., Parigi, B.M., Rochet, J., 2000. Systemic Risk, Interbank Relations and Li-
quidity Provision by the Central Bank. Journal of Money, Credit and Banking 32(3), 611–
638.
Gai, P., Kapadia, S., 2007. Contagion in Financial Networks. Working paper, Bank of
England.
Gompers, P.A., 1995. Optimal Investment, Monitoring, and the Staging of Venture Capi-
tal. Journal of Finance 50(5), 1461–1489.
Gulati, R., 1999. Network Location and Learning: The Influence of Network resources
and Firm Capabilities on Alliance Formation. Strategic Management Journal 20(5), 397–
420.
Gupta, A.K., Sapienza, H.J., 1992. Determinants of Venture Capital Firms’ Preferences
Regarding the Industry Diversity and Geographic Scope of Their Investments. Journal of
Business Venturing 7 (5), 347–362.
Hart, O.D. and J. Moore (1999): “Foundations of Incomplete Contracts,” Review of Eco-
nomic Studies, 66, 115–138.
Hellmann, T., Lindsey, L., Puri, M., 2008. Building Relationships Early: Banks in Ven-
ture Capital. Review of Financial Studies 21(2), 513–541.
Hochberg, Y.V., Ljungqvist, A., Lu, Y., 2007. Venture Capital Networks and Investment
Performance. Journal of Finance 62(1), 251–301.
Page 42 of 48
Hochberg, Y.V., Ljungqvist, A., Lu, Y., 2010. Networking as a Barrier to Entry and the
Competitive Supply of Venture Capital. Journal of Finance 65(3), 829–859.
Huang, H., Xu, C., 2003. Financial Syndication and R&D. Economics Letters 80(2),
141–146.
Jackson, M.O., 2005. A Survey of Models of Network Formation: Stability and Effi-
ciency. Group Formation in Economics: Networks, Clubs and Coalitions, edited by De-
mange, G., Wooders, M., Cambridge University Press, 11–88.
Jackson, M.O., 2009. An Overview of Social Networks and Economic Applications. The
Handbook of Social Economics, edited by Benhabib, J., Bisin, A., Jackson, M.O., Elsevier
Press, Forthcoming.
Lagunoff, R., Schreft, S.L., 2001. A Model of Financial Fragility. Journal of Economic
Theory 99(1-2), 220–264.
Leitner, Y., 2005. Financial Networks: Contagion, Commitment, and Private Sector
Bailouts. Journal of Finance 60(6), 2925–2953.
Leone, A.J., Rock, S., Willenborg, M., 2007. Disclosure of Intended Use of Proceeds and
Underpricing in Initial Public Offerings. Journal of Accounting Research 45(1), 111–153.
Lerner, J., 1994. The Syndication of Venture Capital Investments. Financial Manage-
ment 23(3), 16–27.
Myers, S., Majluf, N.S., 1984. Corporate Financing and Investment Decisions When
Firms Have Information That Investors Do Not Have. Journal of Financial Economics 13(2),
187–221.
Nahata, R., 2008. Venture Capital Reputation and Investment Performance. Journal of
Financial Economics 90(1), 127–151.
Norton, E., Tenenbaum, B.H., 1993. Specialization versus Diversification as a Venture
Capital Strategy. Journal of Business Venturing 8 (5), 431–442.
Ozdemir, S.Z. (2006). To Go Solo or To Syndicate: Determinants of Tie Formation in
The U.S. Venture Capital Industry. Working Paper, University of Chicago.
Pichler, P., Wilhelm, W., 2001. A Theory of the Syndicate: Form Follows Function.
Journal of Finance 56(6), 2237–2264.
Piotroski, J., 1999. The Impact of Reported Segment Information on Market Expecta-
tions and Stock Prices. Working Paper, University of Chicago.
Page 43 of 48
Sah, R.K., Stiglitz, J.E., 1986. The Architecture of Economic Systems: Hierarchies and
Polyarchies. American Economic Review 76(4), 716–727.
Seppä, T., Jääskeläinen, M., 2003. How the Rich Become Richer in Venture Capital:
Firm Performance and Position in Syndication Networks. Working Paper, Helsinki Univer-
sity of Technology.
Smith, D.G., 2001. How Early Stage Entrepreneurs Evaluate Venture Capitalists. Work-
ing Paper, Law of Lewis & Clark College.
SØrensen, M., 2007. How Smart Is Smart Money? A Two-sided Matching Model of Ven-
ture Capital. Journal of Finance 62(6), 2725–2762.
Sorenson, O., Stuart, T.E., 2001. Syndication Networks and the Spatial Distribution of
Venture Capital Investments. American Journal of Sociology 106(6), 1546–1586.
Sorenson, O., Stuart, T.E., 2008. Bringing the Context Back In: Settings and the Search
for Syndicate Partners in Venture Capital Investment Networks. Administrative Science
Quarterly 53(2), 266–294.
Tykvová, T., 2007. Who Chooses Whom? Syndication, Skills and Reputation. Review of
Financial Economics 16(1), 5–28.
Wang, S., 2009. Convertibles in Sequential Financing. Review of Finance 13(4), 727–
760.
Wang, S. and Hailan Zhou (2004): “Staged Financing in Venture Capital: Moral Hazard
and Risks,” Journal of Corporate Finance, 10 (1), 131–155.
White, H., 1980. A Heteroskedasticity-Consistent Covariance Matrix Estimator and A
Direct Test for Heteroskedasticity. Econometrica 48(4), 817–838.
Wilson, R., 1968. The Theory of Syndicates. Econometrica 36(1), 119–132.
Wright, M., Lockett, A., 2003. The Structure and Management of Alliances: Syndication
in the Venture Capital Industry. Journal of Management Studies 40(8), 2073–2102.
Page 44 of 48
Tables
Table 1. The Sample Distribution
The sample consisted of 40,395 VC investment rounds led by 2,373 lead VCs and made in 15,264 portfolio companies that received their initial VC funding during
1985–2005 and for which there were relevant data in the database. The table presents the distribution of VC investment rounds by funding year and the industry,
location and development stage of the portfolio company. The number of observations and the corresponding percentages are listed.
Year Obs. Percent Industry Obs. Percent Location Obs. Percent Stage Obs. Percent
1985 456 1.13 Agriculture/Forestry/Fishing 70 0.17 California 14863 36.79 Seed Stage 5500 13.62
1986 732 1.81 Biotechnology 2776 6.87 Massachusetts 4605 11.40 Early Stage 10651 26.37
1987 1056 2.61 Business Services 773 1.91 Texas 2358 5.84 Expansion Stage 18025 44.62
1988 1122 2.78 Communications 4565 11.30 New York 1823 4.51 Later Stage 6219 15.40
1989 1211 3.00 Computer Hardware 1901 4.71 Pennsylvania 1335 3.30 Total 40395 100
1990 1227 3.04 Computer Other 105 0.26 Washington 1224 3.03
1991 1030 2.55 Computer Software 9280 22.97 Colorado 1203 2.98
1992 1081 2.68 Construction 98 0.24 New Jersey 1184 2.93
1993 936 2.32 Consumer Related 2252 5.57 Georgia 1059 2.62
1994 886 2.19 Financial Services 630 1.56 Virginia 1057 2.62
1995 1206 2.99 Industrial/Energy 1650 4.08 Illinois 943 2.33
1996 1664 4.12 Internet Specific 7396 18.31 Maryland 901 2.23
1997 2153 5.33 Manufacturing 422 1.04 North Carolina 818 2.03
1998 2690 6.66 Medical/Health 5193 12.86 Minnesota 808 2.00
1999 3966 9.82 Semiconductor/Electronics 2929 7.25 Florida 805 1.99
2000 5061 12.53 Transportation 288 0.71 Connecticut 758 1.88
2001 3576 8.85 Utilities 26 0.06 Ohio 591 1.46
2002 2593 6.42 Other 41 0.10 Oregon 401 0.99
2003 2465 6.10 Total 40395 100 Tennessee 400 0.99
2004 2619 6.48 Arizona 345 0.85
2005 2665 6.60 other 2914 7.21
Total 40395 100 Total 40395 100
Page 45 of 48
Table 2. Variable Definitions, Measures and Data Sources
The table presents definitions and measures for the dependent, independent and control variables. Dummy
variables are indicated by *. Variables used as the natural logarithm are indicated by **.
Variables Definition and Measurement
Dependent Variables
Syndicate Size** The logarithm of the number of VCs participating in an investment round.
Network Degree The normalized number of VCs that a lead VC had syndicated with in the 5 years prior to the current investment round.
Network Outdegree The normalized number of VCs that a lead VC had invited into her syndicates in the 5 years prior to the current investment round.
Network Indegree The normalized number of VCs that a lead VC had been invited into their syndicates in the 5 years prior to the current investment round.
Investment Risk:
Company Age** The logarithm of the age of a portfolio company at the current investment round. The age of a portfolio company is the number of years between its founding and the year of the current investment round.
Industry Distance (%) The proportion of investment rounds made previously by a lead VC that were not made in the given portfolio company’s industry.
Stage Distance (%) The proportion of investment rounds made previously by a lead VC that were not made in the same development stage as the given portfolio company.
Geographic Distance** The logarithm of the geographic distance (in kilometers) between the resident state of a portfolio company and its lead VC.
Portfolio Company Quality:
Total VC Funding** The logarithm of one plus the total VC funding across all financing rounds.
Quality Indicator* A dummy variable indicating whether the VC-backed portfolio company went public or was acquired before the end of 2009.
Lead VC Type:
Corporate VC Indicator* A dummy variable indicating whether a lead VC firm was affiliated with a corporation.
Institutional VC Indicator* A dummy variable indicating whether a lead VC firm was affiliated with an institution.
Government VC Indicator* A dummy variable indicating whether a lead VC firm was affiliated with a government.
Page 46 of 48
Table 3. Summary Statistics of VC Investment Rounds
The table presents summary statistics describing 40,395 VC investment rounds made in those portfolio
companies that received initial VC funding during 1985–2005 and for which there were relevant data in the
database. The definitions, measures and data sources of the variables are described in Table 2. The quartiles,
means, standard deviations and the number of observations are presented.
Obs. Mean Std. 0.25 0.5 0.75
Dependent Variables:
Syndicate Size 40,395 2.37 1.89 1 2 3
Network Degree (%) 40,395 6.91 7.48 1.23 4.39 9.82
Network Outdegree (%) 40,395 3.02 4.71 0.22 1.16 3.62
Network Indegree (%) 40,395 1.34 1.37 0.25 0.90 2.08
Investment Risk:
Company Age (years) 40,395 4.79 8.43 1 3 5
Industry Distance (%) 40,395 79.85 20.53 73.01 85.17 94.25
Stage Distance (%) 40,395 67.54 20.55 54.44 70 81.90
Geographic Distance (km) 40,395 1207.95 1457.47 0 363.9 2505.9
Portfolio Company Quality:
Total VC Funding ($mil) 40,395 31.08 40.24 6.44 18.11 41.22
Quality Indicator 40,395 0.38 0.48
Lead VC Type:
Corporate VC Indicator 40,395 0.03 0.17
Institutional VC Indicator 40,395 0.04 0.20
Government VC Indicator 40,395 0.11 0.31
Page 47 of 48
Table 4. Correlation Matrix
The table presents the Spearman and Pearson correlation matrices among all the variables respectively in the upper and lower diagonals. The sample consisted of
40,395 VC investment rounds made in those portfolio companies that received initial VC funding during 1985–2005 and for which there were relevant data in the
database. The definitions, measures and data sources of the variables are described in Table 2. The significance level at 5% is indicated by *.
1 2 3 4 5 6 7 8 9 10 11 12 13
Dependent Variables:
1 Syndicate Size 0.23* 0.29* 0.15* -0.03* -0.05* 0.01 0.07* 0.41* 0.14* -0.01 0.01 -0.06*
2 Network Degree 0.20* 0.88* 0.90* -0.04* -0.04* -0.02* 0.10* 0.14* 0.14* -0.11* 0.03* -0.14*
3 Network Outdegree 0.24* 0.86* 0.71* -0.01* -0.04* -0.05* 0.12* 0.23* 0.13* -0.13* -0.003 -0.15*
4 Network Indegree 0.12* 0.86* 0.58* -0.05* -0.06* -0.004 0.05* 0.12* 0.12* -0.11* -0.002 -0.13*
Investment Risk:
5 Company Age -0.03* -0.06* -0.03* -0.06* -0.04* -0.13* 0.08* -0.03* 0.01* -0.003 0.05* 0.01
6 Industry Distance 0.002 0.16* 0.14* 0.13* 0.002 0.15* 0.001 -0.10* -0.02* -0.04* 0.04* 0.04*
7 Stage Distance 0.02* 0.06* 0.04* 0.06* -0.14* 0.120* -0.05* -0.01 0.03* -0.002 -0.02* -0.004
8 Geographic Distance 0.05* 0.08* 0.11* 0.03* 0.10* 0.002 -0.04* 0.09* 0.02* 0.04* 0.08* -0.06*
Portfolio Company Quality:
9 Total VC Funding 0.37* 0.08* 0.11* 0.09* -0.08* -0.07* -0.03* 0.06* 0.16* 0.04* -0.02* -0.12*
10 Quality Indicator 0.15* 0.14* 0.12* 0.11* 0.003 -0.001 0.02* 0.004 0.16* 0.007 -0.003 -0.06*
Lead VC Type:
11 Corporate VC Indicator -0.004 -0.09* -0.09* -0.06* -0.007 -0.08* 0.004 0.04* 0.02* 0.01* -0.07* -0.04*
12 Institutional VC Indicator 0.02* 0.07* 0.05* 0.06* 0.067* 0.03* -0.02* 0.10* -0.04* -0.003 -0.07* -0.06*
13 Government VC Indicator -0.06* -0.11* -0.09* -0.11* 0.01* 0.02* 0.01* -0.06* -0.14* -0.06* -0.04* -0.06*
Page 48 of 48
Table 5. Regression Results
The sample consisted of 40,395 VC investment rounds made in those portfolio companies that received initial
VC funding during 1995–2005 and for which there were relevant data in the databases. The table presents OLS
regression results relating the dependent variable (Syndication Size, Network Degree, Network Outdegree or
Network Indegree) to the factors including investment risk and portfolio company quality as well as other control
variables. The definitions, measures and data sources of the variables are described in Table 2. Funding year,
industry and round sequence fixed effects were included in all the regressions (not reported). Intercepts are not
reported. Robust t-values are in parentheses. The significance levels at 1%, 5% and 10% are identified by ***, ** and
*, respectively.
Syndicate Size Network Degree
Network Outdegree
Network Indegree
Expected Sign
1 2 3 4 5
Investment Risk:
Company Age - -0.064 -0.022 -0.582 -0.275 -0.108
(-17.40)*** (-6.00)*** (-13.60)*** (-10.05)*** (-12.66)***
Industry Distance + 0.002 0.002 0.053 0.031 0.009
(11.32)*** (10.65)*** (44.39)*** (43.49)*** (37.64)***
Stage Distance 0.001 0.001 0.006 0.002 0.002
(5.54)*** (4.68)*** (4.07)*** (2.88)*** (5.33)***
Geographic Distance + 0.009 0.006 0.102 0.115 0.004
(10.73)*** (6.97)*** (12.00)*** (20.29)*** (2.37)**
Portfolio Company Quality:
Total VC Funding + 0.11 0.599 0.354 0.083
(54.89)*** (27.61)*** (27.56)*** (19.54)***
Quality Indicator + 0.126 0.684 0.468 0.085
(19.13)*** (9.94)*** (9.90)*** (6.26)***
Lead VC Type:
Corporate VC Indicator +/- -0.046 -0.039 -2.032 -1.637 -0.253
(-2.75)*** (-2.35)** (-12.67)*** (-26.3)*** (-5.91)***
Institutional VC Indica-tor
+/- 0.057 0.061 1.174 0.393 0.232
(5.48)*** (6.03)*** (9.03)*** (4.29)*** (8.04)***
Government VC Indica-tor
+/- -0.16 -0.023 -2.819 -1.411 -0.554
(-9.89)*** (-1.41) (-25.42)*** (-23.93)*** (-23.55)***
Year, industry and round sequence fixed effects
Present Present Present Present Present
No. of Observations 40,395 40,395 40,395 40,395 40,395
R2 0.115 0.197 0.295 0.178 0.172